Skip PIP-screened regions gracefully instead of erroring

[Yining97]

What

summaryStatsQc(pipCutoffToSkip) empties a no-signal region (df[FALSE, ]) and records qcInfo$entryAudit[[i]]$pipScreenSkipped. Nothing consumed that flag, so fineMappingPipeline looped the 0-variant entry and .fmExtractZN threw "entry has no SNP mcol". Three problems: (1) misleading message (reads like corrupt input); (2) hard stop() → non-zero exit → shows as FAILED in an array job, indistinguishable from a real error; (3) pipScreenSkipped written and never read.

Now fineMappingPipeline (GwasSumStats + QtlSumStats) consults .fmEntrySkipInfo(data, i) before extracting z/N: a screened (or 0-variant) entry is skipped with a clear region skipped: <reason> message and no row; an all-screened collection returns a valid 0-row result (allowEmpty on the builders) instead of erroring.

Impact

• Fixes all three. A PIP-screened region completes as a skip (exit 0, clear reason), not a failure.
• No protocol change (verified): fine_mapping.R handles a 0-row result and gwas_rss_plot.R already write_emptys an empty FMR.
• Non-screened regions unchanged.

Testing

susieR 0.16.6: test_fineMappingPipeline (474) + test_sumstatsQc (825) green — new tests: screened GWAS → empty FMR + skip message; mixed screened+real → only real rows; 0-variant entry skipped without the flag; QtlSumStats screened trait → empty result. Real chr21 AD_Bellenguez: forced screen (cutoff 0.163 > max SER PIP 0.153) → fineMappingPipeline returns a 0-row GwasFineMappingResult with the skip message, where pre-change it errored "entry has no SNP mcol".