Implement fast vectorized reading for transient PSF files, test shows…

[circuitmuggle]

… x37 faster tran.tran.tran file reading without cache file
original read speed of 3.6GB file

optimized read speed of the same 3.6GB file (rm tran.tran.tran.cache before test)

compatible with original flow, optimized with

1. Bulk File Reading (I/O Optimization), Reads the entire file content into memory at once using f.read()
2. Vectorized Parsing (CPU Optimization), Uses numpy to convert the entire array of string tokens to floats in a single C-level operation (np.array(tokens, dtype=float))
3. Simplified "Fast Path" Logic. Original uses a complex, robust parser (likely ply based) that handles every possible edge case and syntax variation, which is computationally expensive.Implemented a _try_fast_read method that assumes the VALUE section (where the heavy data lives) has a regular structure. It directly splits the string and reshapes the array, bypassing the complex parsing logic for the bulk of the data.
4. Hybrid Compatibility, The code now attempts the Fast Path first. If the file structure is complex (e.g., DC operating points or non-standard formats) or if the fast read fails for any reason, it automatically falls back to the original parser.

[KenKundert]

A remarkable idea. Thanks! But it seems that it is failing the tests: the sweep units were not picked up.

[circuitmuggle]

Hi Ken, I modify the code based on the tox tests feedback. It now shows pass the tests. But let me know if I missed something else.

[KenKundert]

I'll let it age for a bit before releasing it to pypi.

It is a very nice enhancement. Thanks!