fix: add nsamples to recognised sigproc header keys

[themavik]

Summary

Fixes #53

Filterbank files that include nsamples in their binary header cause parse_header() to crash with KeyError: 'nsamples' because the key is not in the header_keys dictionary.

Root Cause

In sigpyproc/io/sigproc.py, header_keys defines the set of recognised sigproc header keywords and their struct format codes. The nsamples keyword is a valid sigproc header field (unsigned 32-bit integer), but it was missing from the dictionary.

When parse_header() reads a header key from the file and looks up its format:

key_fmt = header_keys[key]  # KeyError: 'nsamples'

it crashes because nsamples is not a recognised key.

Fix

Added "nsamples": "I" to the header_keys dictionary. This allows the parser to correctly read past the nsamples field in the binary header.

The computed nsamples value on lines 288-290 still overwrites whatever the file declares, preserving the existing behavior of deriving the sample count from the actual file dimensions (datalen, nbits, nchans).

Testing

• Filterbank files without nsamples in the header: behavior unchanged
• Filterbank files with nsamples in the header: parsed correctly instead of crashing
• spp_header print works on files that previously triggered the crash

[pravirkr]

Thanks for the PR. It required extensive refactoring because "nsamples" is a legacy keyword, and we do not want to make it rigid. Also, we need to decide at runtime which nsamples to use: metadata or file length. Solving this in #56. Closing this now.