feat(control): extract text from PDF attachments / 提取 PDF 附件文本

[SivanCola]

Summary

• extract text from referenced PDF attachments before injecting file context
• prefer pdftotext, then fall back to common Python PDF libraries when available
• bound PDF extractor stdout/stderr output and apply kill-tree timeout handling for subprocesses
• preserve prompt-cache stability by keeping extracted PDF text in the per-turn dynamic reference block, not in stable prompts or tool schemas
• return an OCR/multimodal guidance note when text extraction is unavailable or the PDF appears scanned/image-only

Testing

• go test ./internal/control -run 'TestReadFileRef|TestRunPDFTextCommandCapsStderr|TestResolveRefsAttachmentKinds|TestFileRefLine|TestParseRefTokens|TestClassifyRef'
• HOME-isolated go test ./internal/control
• go test ./internal/proc
• git diff --check

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 0b2a69aa89

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[esengine]

Solid — @-referencing a PDF now extracts its text (pdftotext, then a Python fallback), with a context-timeout + process-tree kill + bounded buffers, args passed directly to exec (no shell), and a clean message pointing at OCR/vision for scanned PDFs. Testable seam + refs_test.go. Thanks!