Correctness bugs in benchmark/query measurement path

[riasc]

From a Python audit. Four correctness bugs, all in the benchmarking measurement path.

• Scaffold filter never matches — [chr for chr in allchroms if "SCF" in allchroms] tests membership in the list, not the name. Should be if "SCF" in chr. Scaffold names collide / miscount. segmeter/simulator.py:68
• bedops arbitrary-pairs branch redirects into a file object, not its path — > {reffiles['ref-srt']} stringifies the NamedTemporaryFile object; use .name (and reuse .name in the following bedops call). Writes to a garbage path. segmeter/calls.py:171-175
• if rss_value: treats the not-found sentinel -1 as valid — get_rss_from_stderr returns -1 on miss, which is truthy, so mem becomes a small negative instead of 0. Use if rss_value > 0:. segmeter/calls.py:35
• Untimed / unchecked post-process subprocess — the bedtk dedup subprocess.run(call, shell=True) bypasses tool_call timing and ignores the exit code. Route through tool_call or at least pass check=True. segmeter/calls.py:245, segmeter/calls.py:263