Fix UnicodeDecodeError When Reading Job Log Files

[AryanAhadinia]

Problem

Reading log.out / log.err from Slurm and FirecREST jobs could raise a UnicodeDecodeError in two scenarios:

• The file is read mid-flush, leaving a truncated multi-byte UTF-8 sequence at the end
• The log contains other malformed byte sequences (e.g. from terminal escape codes or non-UTF-8 tool output)

Example error:

UnicodeDecodeError: 'utf-8' codec can't decode byte 0xe2 in position 232889: invalid continuation byte

Fix

Adds a decode_log(data: bytes) -> str utility that:

1. Strips any trailing incomplete multi-byte UTF-8 sequence (handles mid-flush reads cleanly, without replacement characters)
2. Falls back to errors="replace" for any other malformed bytes elsewhere in the content

Both SlurmLauncher and FirecRESTLauncher now read log files as bytes and pass them through decode_log.

[AryanAhadinia]

The Lightweight CI / Tests (Python 3.12) (pull_request) is normal due to the ongoing efforts in #36.

[AryanAhadinia]

The Lightweight CI / Tests (Python 3.12) (pull_request) is normal due to the ongoing efforts in #36.

Now fixed.