align Mahalanobis and DA PTP with manuscript; harmonize use_empirical_variance default; bump to 0.8.0

[settylab-dotto-bot]

Brings the implementation into agreement with the manuscript on
three statistical points and harmonizes a default-argument
inconsistency that had crept in across deprecated wrappers. Targets
v0.8.0.

These are scalar recalibrations, not changes to what is detected.
The DE Mahalanobis change is a single uniform multiplicative factor
(/√2 on the distance): it does not alter significance detection
or gene ordering — every gene's distance scales by the same constant,
the null scales identically, so relative rankings and FDR call-sets are
unchanged after recalibration. The DA PTP change is likewise a scalar
(a halving, i.e. a constant shift in log space): it leaves the ordering
of points intact and only makes a fixed default threshold slightly
more sensitive. Absolute thresholds tuned against 0.7.0 should be
re-checked, but no qualitative result moves.

Mahalanobis denominator — sum, not average

The manuscript defines the gene-wise distance as
D(a,b) = sqrt((μ_a − μ_b)^T (Σ_a + Σ_b)^(-1) (μ_a − μ_b)),
but the GP-posterior branch in
kompot/differential/differential_expression.py:629 was computing
combined_cov = (cov1 + cov2) / 2. This change drops the /2 so the
combined covariance equals the variance of the difference of two
independent posterior estimators. The sample-variance branch at the
same call site was already correctly summed — which is what made the
inconsistency visible.

Effect — purely a scalar. D contracts by √2 (and D² by 2)
everywhere. Because the identical factor applies to both the
observed and the null distances, this is a uniform rescaling of the
axis: it does not change which genes are called significant and it
does not change gene order. FDR thresholds re-calibrate against the
rescaled null, so call-sets are essentially unchanged; only the raw
numeric magnitude of D differs.

DA posterior tail probability — one-sided

The manuscript prose describes a one-sided test ("the probability that
the true density difference has the opposite sign to the estimated
change"), and the formula
PTP(x_i) = Φ(−|Δ(x_i)| / sqrt(σ_a² + σ_b²)) is one-sided. The
implementation at
kompot/differential/differential_abundance.py:607-610 was emitting
2 · Φ(−|z|) (two-sided). This change removes the spurious
+ np.log(2) so ln_ptp = log Φ(−|z|).

Effect — a scalar shift, slightly more sensitive default. Numeric
PTPs are halved (a constant multiplicative factor; neg_log10_fold_change_ptp
shifts up by log10(2) ≈ 0.301). Point ordering is untouched. The
practical consequence is only that a fixed PTP cutoff becomes a touch
more permissive: a PTP < 1e-3 threshold that meant |z| ≥ 3.29
(two-sided) now means |z| ≥ 3.09 (one-sided) — i.e. the default
threshold is slightly more sensitive. Re-tune any ptp_threshold
calibrated against 0.7.0 if exact parity is required.

use_empirical_variance — False everywhere

The manuscript states empirical variance is disabled by default. This
was already true for the recommended kompot.de() path (via
GPSettings(use_empirical_variance=False)) and for
de_config_template.yaml. Several other publicly-exposed entry points
were inconsistent (default True):

• kompot.differential.DifferentialExpression.__init__
• kompot.differential.expression_model.ExpressionModel.__init__
• kompot.anndata.differential_expression.compute_differential_expression
  (deprecated wrapper)
• kompot.anndata.smooth.compute_smoothed_expression (deprecated wrapper)
• kompot.anndata.smooth.smooth_expression (gp=None fallback)
• kompot/cli/templates/smooth_config_template.yaml

All now default to False, matching the recommended path and the
manuscript. Docstrings and the resource-estimation comment for
combined-covariance memory are brought in line. Code relying on
empirical variance must now opt in explicitly via
GPSettings(use_empirical_variance=True) / the keyword arg.

0.7.0 → 0.8.0

pyproject.toml and kompot/version.py bumped to 0.8.0; a new
CHANGELOG.md section under ## [0.8.0] records the three behavior
changes above with migration notes. No release tag is cut by this PR.

Tests

tests/test_audit_fixes.py adds three focused regression suites pinning
the post-0.8.0 invariants:

• Mahalanobis denominator — monkeypatches
  compute_mahalanobis_distances to capture the covariance kwarg from
  a controlled-kernel fit (cov1 = 2I, cov2 = 3I) and asserts equality
  with 5 · I. Pre-0.8.0 would have produced 2.5 · I.
• DA PTP — replicates the exact ln_ptp formula and asserts it
  equals the closed-form Φ(−|z|), both directly and through a full
  DifferentialAbundance.fit().predict() round-trip.
• use_empirical_variance default — introspects every public entry
  point's signature and asserts the default is False; also parses both
  CLI YAML templates.

tests/test_empirical_variance.py — the one pre-existing test pinning
the old buggy default (test_default_is_on) is flipped and renamed
test_default_is_off. The Mahalanobis-approach tests
(tests/test_mahalanobis_approaches.py) compare implementations against
each other rather than pinned numerics, so they are insensitive to the
scale shift.

[settylab-dotto-bot]

Added kompot.plot.lollipop to this branch (commit ed30264) — an
ax-embeddable gene-set-enrichment lollipop ported from the manuscript
(Fig 3 G/L) and generalized.

One row per enriched term; a stem runs to a dot whose x-position encodes
significance (-log10(FDR) by default, or any score column) and whose area
encodes the matched-gene count, with a dashed FDR = 0.05 guide and an
in-axes aesthetic key. Like dotplot, it composes into an externally-provided
ax= rather than building its own GridSpec.

Headline feature is input flexibility: accepts a kompot.plot.StringDBReport
(its get_functional_enrichment() is called for you), the signal-sorted
frame that method returns, or a generic enrichment table from another tool
(gseapy/enrichr, GOATOOLS, clusterProfiler) via term_col / score_col /
count_col / fdr_col mapping params with case-insensitive autodetection —
including the gseapy "k/K" Overlap-string count parser.

Additive only; no existing plot code touched. New file
kompot/plot/lollipop.py + tests/test_plot_lollipop.py (22 tests, green),
exported from kompot.plot, with CHANGELOG + docs updates.

Synthetic-data example renders and details are tracked internally.