[Bug] scannum with a bound variable returns each scan once per matching candidate

[hsingh84ai]

What happens

A query that binds a variable and returns scannum reports the same scan many times over —
once per candidate mass that matched it — so the reported match count is a large multiple of
the number of scans that actually matched.

Measured on a 48-spectrum Thermo file with 14 MS1 scans:

QUERY scannum(MS1DATA) WHERE MS1MZ=X

returns 2,783 rows over those 14 scans. The distinct set of scan numbers is correct; it
is the multiplicity that is wrong.

Why

A variable query is executed once per candidate mass and the per-candidate results are
concatenated (msql_engine.py, the comment column is set to the candidate's m/z at
line ~342). The de-duplication that follows is guarded on that column:

if "comment" in collated_df:
    collated_df["truncated_comment"] = collated_df["comment"].astype(float).astype(int)
    collated_df = collated_df.drop_duplicates(subset=["scan", "truncated_comment"])

Two things combine:

1. the subset is ["scan", "truncated_comment"], so it de-duplicates per (scan, candidate)
   rather than per scan — by design, since a scaninfo result wants one row per candidate; and
2. the scannum result frame is rebuilt from scratch and carries only a scan column
   (msql_engine.py:606-614), so "comment" in collated_df is False and no de-duplication
   runs at all.

So for scannum specifically, every candidate that matched a scan contributes another copy
of that scan.

Suggested fix

scannum's contract is a list of scan numbers, so de-duplicating it unconditionally on
scan after collation would match what the function already does within a single execution
(list(set(...)) at line 610/612).

Note

scaninfo is not affected in the same way — its frame does carry comment.