feat(state): persist context_length for dashboard cap fallback

[OmarB97]

Summary

Persist context_length to state.db on session init. Downstream
dashboards (notably MeshBoard's context chip) can render a
definitive context-window cap even when their static
model-metadata tables and live HTTP probes can't answer for an
unrecognized backend.

Motivation

The chip computes its cap via:

1. A static model-metadata table (context_window_hint(model_name))
2. A live HTTP probe of the loaded backend's /slots / /api/show
   / /v1/models endpoints

For unrecognized models behind unreachable backends, neither answers
and the chip degrades to ctx N / ? with no bar. Hermes knows the
answer authoritatively via self.context_compressor.context_length
— it just wasn't surfaced.

Changes

hermes_state.py:

• SCHEMA_SQL: new context_length INTEGER DEFAULT 0 column on the
  sessions table. Declarative reconciliation picks it up.
• New set_context_length(session_id, length) method — simple SET
  with INSERT-OR-IGNORE row guard.

run_agent.py:

• _ensure_db_session: after create_session succeeds, snapshot
  the model's context_length. Constant for the session lifetime;
  written once. Best-effort try/except.

Tests

4 new tests in tests/test_hermes_state.py:

• write succeeds
• zero clears (caller signals 'unknown')
• INSERT-OR-IGNORE creates row if missing
• declarative reconciliation adds the column on existing DBs

Test run

$ python3 -m pytest tests/test_hermes_state.py
============================= 230 passed in 7.67s ==============================

Compatibility

• Existing DBs: declarative reconciliation adds the column on next
  Hermes startup. No migration script.
• create_session signature unchanged. The setter is called as
  a follow-up in _ensure_db_session.
• Dashboard: a MeshBoard PR consumes the new column via the
  existing context_window resolution chain. No coordination
  required — when present, the column wins; when missing
  (older Hermes), the chip falls back to its existing chain.

[github-actions]

🔎 Lint report: feat/state-db-context-length vs origin/main

ruff

Total: 0 on HEAD, 0 on base (➖ 0)

🆕 New issues: none

✅ Fixed issues: none

Unchanged: 0 pre-existing issues carried over.

ty (type checker)

Total: 8386 on HEAD, 8386 on base (➖ 0)

🆕 New issues (3):

Rule	Count
invalid-argument-type	3

First entries

run_agent.py:7496: [invalid-argument-type] invalid-argument-type: Argument to function `build_anthropic_client` is incorrect: Expected `str`, found `str | dict[Unknown, Unknown] | Any | ... omitted 3 union elements`
run_agent.py:13782: [invalid-argument-type] invalid-argument-type: Argument to function `_is_oauth_token` is incorrect: Expected `str`, found `str | dict[Unknown, Unknown] | Any | ... omitted 3 union elements`
run_agent.py:13785: [invalid-argument-type] invalid-argument-type: Argument to function `len` is incorrect: Expected `Sized`, found `(str & ~AlwaysFalsy) | (dict[Unknown, Unknown] & ~AlwaysFalsy) | (Any & ~AlwaysFalsy) | ... omitted 3 union elements`

✅ Fixed issues (3):

Rule	Count
invalid-argument-type	3

First entries

run_agent.py:7480: [invalid-argument-type] invalid-argument-type: Argument to function `build_anthropic_client` is incorrect: Expected `str`, found `str | dict[Unknown | str, Unknown | str | dict[str, str]] | Any | ... omitted 3 union elements`
run_agent.py:13769: [invalid-argument-type] invalid-argument-type: Argument to function `len` is incorrect: Expected `Sized`, found `(str & ~AlwaysFalsy) | (dict[Unknown | str, Unknown | str | dict[str, str]] & ~AlwaysFalsy) | (Any & ~AlwaysFalsy) | ... omitted 3 union elements`
run_agent.py:13766: [invalid-argument-type] invalid-argument-type: Argument to function `_is_oauth_token` is incorrect: Expected `str`, found `str | dict[Unknown | str, Unknown | str | dict[str, str]] | Any | ... omitted 3 union elements`

Unchanged: 4421 pre-existing issues carried over.

Diagnostics are surfaced as warnings — this check never fails the build.