Order of logpdf evaluation in BPINNs is incorrect

[affans]

Describe the bug 🐞

Suppose you have 4 ODE parameters and for each one, there is a prior assigned:

invpriors= [        
        Normal(1, 2), 
        Normal(2, 2), 
        Normal(2, 2),
        Normal(0, 2)]

Also suppose you have a vector of parameters θ = [weights..., biases..., λ1, λ2, λ3, λ4] where λ1, λ2, λ3, λ4 are parameter values of the ODE system. The function priorweights() on line 247 of advancedHMC_MCMC.jl evaluates the log PDF at these parameter values:

invlogpdf = sum(
        logpdf(invpriors[length(θ) - i + 1], θ[i])
    for i in (length(θ) - ltd.extraparams + 1):length(θ))

Except the indices are incorrect. Unrolled, this will evaluate param[4](λ1), param[3](λ2), param[2](λ3) and param[1](λ4). In many cases this is probably not a big deal, but still could be fixed quickly.

Expected behavior
Should evaluate param[1](λ1), param[2](λ2), param[3](λ3) and param[4](λ4). The fix seems easy: Replace

 logpdf(invpriors[length(θ) - i + 1], θ[i])

with

logpdf(invpriors[i - length(θ) + ltd.extraparams], θ[i])

or some other cleaner version. Is this worth fixing? Happy to submit a quick PR.

[affans]

A cleaner solution is:

ode_param_values = θ[length(θ) - ltd.extraparams + 1:end]
invlogpdf = sum(logpdf.(invpriors, ode_param_values))