Avoid unnecessary cache purges

[ababic]

These changes update the signal handlers so that cache purges for document URLs are only triggered when the "collection" or "file" fields could have been updated (and skip them if not).

For example, if I had a background task that transcribed documents and saved the result with something like...

document.transcription = transcription_text
document.save(update_fields=['transcription'])

...the purge would be skipped.