[Feature]: grep line numbers/file grouping; list_dir honors .gitignore

[shauryagangrade]

Problem

grep output is a raw match dump (no line numbers or file headers), which is
hard for the model and the user to navigate in big codebases. list_dir
also lists build artifacts and __pycache__ instead of the meaningful tree.

Proposed approach

• grep: prefix matches with path:line: and group by file (respecting a
  reasonable result cap).
• list_dir: skip __pycache__, .git, and entries matched by .gitignore
  (a lightweight parser, or respect it when a .gitignore exists).

Where to look

• gcode/tools.py:242-285 (_grep_python), :169-191 (list_dir)

Acceptance criteria

• grep output includes path:line:content; capped so context isn't blown.
• list_dir omits ignored/transient entries.

Difficulty

Low — good first issue.