[AI Agent Test] ply compatible fast read, mainly implementated in parse.py

[circuitmuggle]

Dear Ken:

This pull request is not for merge purpose but It's also clean to merge if you want. (Same x37 speed up and pass all 39 tox tests)

I am testing Gemini 3 Pro/Claude Sonnet 4.5 (in AntiGravity IDE) as AI Agent to write the code. This implementation has the same effect as the previous pull. The only difference is that the fast_read is implemented in tight coupling with parse.py, instead of the fast_read bypass ply approach. Hence the psf.py file is relatively clean.

I would like to use this as s transition study from Computer-Aided Design(CAD) to Human-Aided AI Design (HAAD) :p

I will probability close this pull request after 5 days without action

Regards,
Gaofeng

[KenKundert]

Which version is preferred?

[circuitmuggle]

Hi Ken,

Thank you so much for your feedback.

I prefer this latest version for several reasons:

1. Tighter integration with PLY
2. Less code
3. A cleaner, more elegant architecture
4. A single, unified parsing path that is easier to reason about

The previous version felt more like a workaround that bypassed PLY’s lexer/parser. I would only recommend that approach if:
a) there are concerns about PLY version stability,
b) frequent parser-debugging is needed, or
c) someone is unfamiliar with PLY internals.

I also added additional tests to improve coverage and to ensure all checks pass.