fix: bound regex project-library searches

Problem

search_project_library accepts arbitrary Python regular expressions and evaluates them against every line of every readable library file. There is no pattern, input-size, file-count, or execution-time bound. max_matches only stops after a match is found.

Impact

A catastrophic-backtracking pattern such as (a+)+$ against a long non-matching line can monopolize a shared executor worker. Repeated requests can exhaust the worker pool and degrade unrelated MCP tools.

Suggested fix

Use a timeout-capable regex engine, or bound searched file/line sizes and reject patterns that exceed a defined complexity policy. Add a total scan-byte/file limit independent of max_matches, and return an explicit truncation result when that limit is reached.

Review evidence

• dataiku_mcp/tools/project_libraries.py:343
• User-controlled regex compilation is at lines 364-371.
• It reads each file and evaluates every line at lines 375-406.

Start in dataiku_mcp/tools/project_libraries.py at search_project_library and inspect regex compilation around lines 364-371 and file scanning at lines 375-406. Decide the bounded-search behavior from the suggested timeout or complexity approach, including total scan limits independent of max_matches. Done means untrusted patterns and large scans cannot monopolize workers, and reaching a limit returns an explicit truncation result.