[performance] filter model and database table improvements (#4277)

- removes unnecessary fields / columns (created_at, updated_at)
- replaces filter.context_* columns with singular filter.contexts bit field which should save both struct memory and database space
- replaces filter.action string with integer enum type which should save both struct memory and database space
- adds links from filter to filter_* tables with Filter{}.KeywordIDs and Filter{}.StatusIDs fields (this also means we now have those ID slices cached, which reduces some lookups)
- removes account_id fields from filter_* tables, since there's a more direct connection between filter and filter_* tables, and filter.account_id already exists
- refactors a bunch of the filter processor logic to save on code repetition, factor in the above changes, fix a few bugs with missed error returns and bring it more in-line with some of our newer code

Reviewed-on: https://codeberg.org/superseriousbusiness/gotosocial/pulls/4277
Co-authored-by: kim <grufwub@gmail.com>
Co-committed-by: kim <grufwub@gmail.com>

test/envparsing.sh

@@ -42,6 +42,7 @@ EXPECT=$(cat << "EOF"
     "cache-domain-permission-subscription-mem-ratio": 0.5,
     "cache-emoji-category-mem-ratio": 0.1,
     "cache-emoji-mem-ratio": 3,
+    "cache-filter-ids-mem-ratio": 2,
     "cache-filter-keyword-mem-ratio": 0.5,
     "cache-filter-mem-ratio": 0.5,
     "cache-filter-status-mem-ratio": 0.5,