[sw/sca] Optimize/Clean up code of batch modes in sca serial files

### Description

This issue is related to https://github.com/lowRISC/opentitan/pull/17416

@jadephilipoom made some good points on the batch mode implementation in the ecc256_keygen_serial.c file, which are also valid to each batch mode implementation in the other serial files. Hence, these suggested modifications should be applied to every serial file.

The points are:
- To fully comply with the style guide.
- To remove the global key/message/seed/etc. arrays, which store all values from each batch, and instead allocate enough stack memory locally for a single key/message/seed/etc. value. If we then directly process each single value this would allow us to remove the maximum number of traces per batch and also the second for-loop.
- To use `memcpy(num_traces, data, data_len)` instead of `num_traces = read_32(data)` . This is safer.

Start with ecc256_keygen_serial.c and compare its batch-mode implementation with the corresponding implementations in the other serial files. Review the style-guide concerns, global batch arrays, second loop, and num_traces handling described in the issue. Done means the same cleanup is applied consistently across all affected serial files.

Written by the indexing model from the issue text.