Consider extraction errors as expected operational errors without crashing the engine

[Ndpnt]

Changed

• Consider extraction errors as expected operational errors without crashing the engine

Added

• Added label empty content to reported issues on GitHub when server returns empty content or when PDF content is unextractable
• Added label invalid selector to reported issues on GitHub when CSS selector is invalid

Fixed

• Prevent engine crashes caused by invalid URLs in source document links
• Ensure the CLI option --types is not ignored
• Fix counts of terms when they are specified via the CLI option --types
• Fixed display issues for service ID and terms type associated with errors

---

Other technical improvements:

• Make logger silent in test env
• Also allow passing no params to the track function
• Ensure all errors from fetch function will bubble as FetchDocumentError
• Ensure all errors from extract function will bubble as ExtractDocumentError
• Minor improvements in logged messages

[clementbiron]

It sounds great and, well done!

[MattiSG]

Looks like very impactful and well-thought-out changes! I didn't review src/archivist/extract/index.test.js.

[MattiSG]

Suggested rename: termsWithNoText or termsWithoutText.

[MattiSG]

Assuming typos and copywriting changes are taken into account, LGTM! 👏
So many fixes and improvements, nice! There were a few riders though, such as the terms types filtering and count correction; it's always better to ship them separately 😉