feat: add support for thinking level 'max' (#283)

[kindralai]

Problem

The thinking level currently supports only "" (off), "low", "medium", and "high". Some reasoning models benefit from even more thinking depth. Adding a "max" level gives the model the widest feasible reasoning budget.

The _reasoning_kwargs() method in core/llm.py silently returns {} for any value outside ("low", "medium", "high"), so "max" was being silently ignored.

Solution

• core/llm.py: Add "max" to the valid levels in _reasoning_kwargs() so it passes through as a thinking/reasoning effort value to all providers
• core/agents.py: Add "max" to _LLM_THINKING_LEVELS tuple so per-agent LLM overrides accept it
• core/config.py: Update all 8 thinking_level field doc comments to document | "max"
• UI (agent_editor.html, base.html): Add "Max" option to the per-agent dropdown and to the global LLM config datalist
• Docs (subagents.mdx): Update valid values

Acceptance criteria

• LLMClient("anthropic", "x", thinking_level="max")._reasoning_kwargs() returns {"thinking": {"type": "adaptive"}, "output_config": {"effort": "max"}}
• LLMClient("openai", "x", thinking_level="max")._reasoning_kwargs() returns {"reasoning_effort": "max"}
• _as_llm_config({"thinking_level": "max"}) returns {"thinking_level": "max"}
• All 8 config field doc comments updated
• "Max" option available in per-agent LLM override dropdown
• "Max" listed in the datalist for the global LLM config thinking level input

Closes #283

[mattmezza]

fix linting - CI fails

[kindralai]

Pushed fix — shortened the line to stay under 100 chars. CI re-running now.

[kindralai]

CI green on the linting fix (commit 922a4fb). The changes requested by @mattmezza are addressed — requesting re-review.

[kindralai]

@mattmezza CI green — ready for re-review