fix: skip system context for coding session auto-names

[danshapiro]

Summary

• root cause: Codex/Claude parsers stored system-context user records like AGENTS.md as firstUserMessage before applying the system-context filter
• fix: add shared user-authored text extraction and use it for provider parsing plus lightweight indexer titles
• tests: cover AGENTS-only records, IDE request extraction, and parser firstUserMessage behavior

Verification

• npm run test:vitest -- test/unit/server/coding-cli/utils.test.ts test/unit/server/coding-cli/codex-provider.test.ts test/unit/server/coding-cli/claude-provider.test.ts test/unit/server/coding-cli/session-indexer.test.ts --run
• npm run check

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: cacd42fd88

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Preserve full IDE request text for firstUserMessage

When an IDE-context record contains a multi-line ## My request for Codex: section, this returns extractFromIdeContext(...), which only returns the first non-empty line of that section. The same userAuthoredText is now used to populate firstUserMessage in both Claude and Codex providers, so sessions with requests like Implement X\n- handle A\n- handle B lose everything after the first line; that truncated value is later used for AI title generation, session search, and first-chat exclusion filters. Title extraction can still use the first line, but the canonical first user message should keep the full request body.

Useful? React with 👍 / 👎.