Strangler Stage 2: tighten to industry-standard complexity thresholds

[sergio-sisternes-epam]

Context

PR #1464 (Stage 1, issue #1077) lowered Ruff complexity thresholds from their initial ceilings to intermediate values. Stage 2 should continue ratcheting towards industry-standard McCabe thresholds.

Stage 2 targets

Metric	Stage 1 (current)	Stage 2 target	Industry standard
max-complexity	50	<=20	10-15 (ruff, httpie)
max-branches	60	<=25	12 (default)
max-statements	200	<=100	50 (default)
max-args	15	<=10	5 (default)
max-returns	12	<=8	6 (default)

Scope

1. Split the 6 files currently over 1400 lines (deferred from Stage 1, MAX_LINES set to 2000 as a compromise):

   • Identify natural module boundaries in each file
   • Extract cohesive sub-modules with clean import graphs

2. Decompose remaining complex functions to meet Stage 2 thresholds

3. Replace IntegratorBundle Any-typed fields with a BaseIntegrator protocol (panel nit from PR refactor: Strangler Stage 1 -- tame worst complexity outliers #1464)

Follow-up from PR #1464 panel review

• Python Architect: type IntegratorBundle fields properly
• OSS Growth Hacker: max-complexity=50 is still 2x contributor-friendly threshold
• CLI Logging Expert: align _clean_toml_mcp_config output contract with siblings (partially addressed)

References

• Parent: Strangler-fig: progressively tighten Ruff complexity thresholds #1076
• Stage 1: Strangler Stage 1: tame the worst outliers #1077, PR refactor: Strangler Stage 1 -- tame worst complexity outliers #1464