add bulletproofing for the case that waveform data come before the first channel header word

[tomjunk]

Description

This PR _Fixes #120 _

The method decode_data in NevisTPCUtilities.cc is not guarded against the case where data that are interpreted as ADC data (either with type kADC or kADCHuffman) come before the first kChannelHeader word in a fragment. Such data are corrupt, as we do not know what channel they correspond to, and the currentWaveformPtr is uninitialized, causing a segfault. This was spotted trying to decode data from run 18101, taken January 7, 2025.

Related Repository Branches

No related branches, just this one.

Testing details

Bug reproduction before this PR:

gdb --args lar -c run_decoders_job.fcl root://fndca1.fnal.gov:1094//pnfs/fnal.gov/usr/sbnd/archive/sbn/sbn_nd/data/raw/unknown/v1_10_04/sbnd_daq_v1_10_04/daq/00/01/81/01/data_EventBuilder1_art1_run18101_1_strmUnknown_20250107T214345.root

Program received signal SIGSEGV, Segmentation fault.
sbndaq::NevisTPCDecoder::decode_data (this=this@entry=0x7ffffffdd0b0, data_ptr=, n_words=n_words@entry=40593, wvfm_map=...)
at /home/trj/saut/triggerdecoder/sbndc4/srcs/sbndaq_artdaq_core/sbndaq-artdaq-core/Overlays/SBND/NevisTPC/NevisTPCUtilities.cc:97
97 currentWaveformPtr->emplace_back( currentWaveformPtr->back() + differences[i_diff - 1] );

Decoding this file succeeds with this PR, but the data do look corrupt from run 18101 Lots of invalid fragment handles, duplicate
channel headers, mismatching headers and trailers, and the event displays look choppy. But the decoder succeeds.

same story with data_EventBuilder3_art1_run18101_1_strmUnknown_20250107T214346.root

I also tested a normal file to make sure the decoding wasn't broken by this PR:

data_EventBuilder5_art1_run18214_100_strmBNBZeroBias_20250205T145349.root

decoded just fine.

[tomjunk]

I also ran indent-region and untabify in emacs so there are a few changes to the whitespace.