Fix Codex profile reasoning_effort and verbosity defaults

[allenzhou101]

Summary

The Codex CLI defaults both model_reasoning_effort and model_verbosity
to "low", but gpt-5.2-codex (the default Codex model) only accepts
"medium" for both. Out-of-the-box codex exec against the AI Gateway
fails with:

Unsupported value: 'low' is not supported with the 'gpt-5.2-codex' model. Supported values are: 'medium'.

The error covers both the reasoning.effort and text.verbosity request
parameters, depending on which the model rejects first.

This PR sets both fields to "medium" in two places:

1. The generated profile config in ~/.codex/default.config.toml —
   ensures the value is captured for any tool that inspects the config.
2. Explicit -c flags on codex exec — CLI flags have the highest
   precedence and we observed the profile-only setting being silently
   overridden by the CLI's "low" default in some Codex versions.

generateCodexConfig now accepts an optional reasoningEffort argument
so callers can still override per-run via
model: "gpt-5.2-codex?reasoningEffort=high".

Test plan

• Unit tests cover "medium" defaults and explicit-override paths
  for both AI Gateway and direct OpenAI configs (6 cases in
  src/lib/agents/codex.test.ts).
• npm test — 184 passed | 14 skipped.
• npm run lint clean.
• End-to-end against the Vercel AI Gateway: verified by linking
  the patched build into a0-local
  and running a one-prompt Codex smoke. Without the fix, turn.failed
  with the Unsupported value: 'low' error within ~10s. With the fix,
  a real response is produced in ~31s.

Notes

• Adds vercel-agent-eval-*.tgz to .gitignore so local npm pack
  artifacts don't leak into commits.
• Stacks well with the recent
  #138 / #139 Codex profile-config fix
  — that PR fixed the legacy profile = "default" rejection; this one
  fixes the next layer (reasoning effort + verbosity defaults) so Codex
  actually completes a turn.