fix(tools): replace ± with +/- in session_search docstring to avoid unicode parse errors

[Morad37]

The session_search tool docstring contains 5 instances of the ± character (U+00B1). Several provider APIs (DeepSeek, OpenRouter) reject this character in tool descriptions, returning HTTP 400:

Failed to parse the request body as JSON: tools[9].function.description: invalid unicode code point

Every request fails as soon as session_search is among the enabled tools, provider-agnostic.

Fix: replace all ± with +/- in docstring comments.

Closes #50959

[alt-glitch]

Note: this branch is a whole-fork push (1,291 files changed, ~33.9k additions / ~129.8k deletions) rather than the focused ± -> +/- docstring change described. As-is it's unreviewable and would delete most of the repo. Please rebase onto current main and open a PR containing only the session_search docstring change for #50959.

[Morad37]

The branch was contaminated because the fork main was behind upstream. Force-pushed a clean branch from upstream main -- the diff now shows only the 5 docstring changes described.

[teknium1]

Thanks for the focused compatibility fix. Current main still has the three runtime-facing U+00B1 occurrences in the registered SESSION_SEARCH_SCHEMA at tools/session_search_tool.py:778, tools/session_search_tool.py:787, and tools/session_search_tool.py:789; the schema is registered at tools/session_search_tool.py:903. The PR diff replaces all five target occurrences with ASCII +/- and introduces no behavioral or schema-shape change.

Automated hermes-sweeper review.