Recognize the MS-DIAL 5 suggested-annotation name prefixes - #784

* Export alignment spectrum provenance links

* Add compact alignment peak ID matrix export

* Stop the provenance sidecar asserting what the run did not establish

Review of the rebased branch found the audit sidecar reporting three things that
were not measurements, all of them invisible to the existing tests because both
tests asserted substrings of a row that had a real source peak, and stopped
before the mz column.

The mz column was read from ChromXsTop.Mz in the in-memory overload and from
Mass in the light overload. On an aligned peak ChromXsTop carries the
chromatogram axis, and the gap filler resets it outright, so the column reported
a sentinel for every member that DID have a source peak and a real value only
for the gap-filled ones. Measured on the FastLC demo: 7879 of 7879 rows with
has_source_peak=true carried mz=-1, and 9992 of 9992 gap-filled rows carried a
plausible m/z. Exactly backwards, in every row. Both overloads now read Mass,
which is also what the .mdalign MZ column uses.

A member with no source peak has no source spectrum either, but the gap filler
leaves the spectrum ids at their default 0 while resetting only the peak ids, so
every gap-filled row published ms1_raw_spectrum_id=0, ms1_raw_spectrum_id_top=0
and ms2_raw_spectrum_id=0. A 0 is a real scan index. In a file whose whole
purpose is provenance, that points an auditor at a spectrum belonging to some
other peak. All five raw-spectrum columns are now withheld when there is no
source peak, in both overloads; the light overload had additionally gated
ms2_raw_spectrum_ids on MS2RawSpectrumID >= 0, which is true for that same 0.

The peak-id columns published -2 for a gap-filled cell while the compact peak-ID
matrix published -1 for the same cell, so two files shipped by one feature
disagreed numerically. They now agree on -1, and the distinction the raw values
carried is preserved by name in a new peak_origin column, valued detected,
gap_filled or absent. That is the same lesson as PR #785: name the state rather
than encode it in a magic number, so a reader does not need the gap filler's
internals to interpret a cell.

Also fix the light-mode QA matrix's "Reference matched" column. The switch to
the file-backed quant accessor made light mode derive it from
RepresentativeLibraryID >= 0, but a suggested annotation carries a library id
too, so a precursor-only or low-score suggestion counted as a reference match in
light mode while normal mode reported false. Normal mode resolves the column
through AnnotationName.IsReferenceMatched, the predicate PR #784 centralised.
The store now persists the representative's own IsReferenceMatched verdict and
the accessor reads that. RecordSize goes 102 to 103; the file is a per-run
temporary spill created by Path.GetTempFileName, so the layout carries no
cross-version obligation.

Two guards so this class of defect cannot return silently. WriteMember now goes
through a WriteRow that refuses a field count the header does not declare, since
the header and the two value lists are three hand-maintained arrays. And the
tests assert whole lines rather than substrings, and cover a gap-filled row and
an unwritten light row in both overloads, which is the case that hid all three
divergences.

Build: Release net48, 0 errors. Tests: MsdialCoreTests 316/316 (311 before, plus
five new), MsdialCoreTestAppTests 13/13 including a new
ReadDetailedAlignmentProvenance test alongside master's ReadLbmAnnotatorPriority
test in the project that actually compiles the rebased ConfigParser.

Verified end to end on the FastLC demo, ingesting the output into the companion
msdial_spectrum_catalog: every member with a source peak now carries a usable
m/z, no member without one carries a scan index, and peak_origin splits the
17871 members into 7879 detected and 9992 gap_filled.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

---------

Co-authored-by: DESKTOP-382ETUR\Hiroshi Tsugawa <htsugawa@go.tuat.ac.jp>
Co-authored-by: Claude Opus 5 <noreply@anthropic.com>