Skip Chromium databases that cannot be read instead of ending the artifact

[abrignoni]

Guards twelve Chromium artifacts so one unreadable database no longer ends the artifact.

• Each of these looped over its matching databases with no guard around the query, so a file that could not be read raised out of the loop and the artifact returned nothing for the whole image, dropping every browser already collected. The file is now logged and skipped.
• SQLite treats a non-empty <db>-journal sidecar as a hot journal and starts a recovery that needs to write, which a mode=ro handle cannot do. pixel3_a12 carries one such file, the Gmail WebView Web Data under data_mirror, whose sidecar is 16 bytes of LevelDB text rather than a rollback journal. Driving get_chromeAutofill over that file plus a healthy Chrome Web Data from the same image raises on main and returns the healthy database's 3 rows on this branch.
• That file is no longer reached in a normal run, since the storage-view dedupe in Dedupe duplicate storage paths in per-file artifacts #1135 selects the data/data spelling of the same database. No registered corpus currently exercises the path, so these are guards against a reachable failure rather than repairs of an observed one.
• Covers chromeAutofill, chrome.py (Web History, Web Visits, Search Terms, Downloads, Keyword Search Terms), chromeCookies, chromeLoginData, chromeMediaHistory, chromeDIPS, chromeNetworkActionPredictor and chromeOfflinePages. chrome.py and chromeMediaHistory route their queries through one helper each. report_file now lists only the files actually read.
• Sample data for both Autofill artifacts re-derived across all 17 registered phone corpora, up from 10 and 12.

Verified with aleapp.py profile runs over every registered Android corpus, main against this branch: no error banners either side and 25,774 rows either side, so the guards change nothing where nothing fails.