chore: merge back delete request when we are done processing all its splits

[sandeepsukhani]

What this PR does / why we need it:
By default, we split delete requests with line filters by 24-hour intervals. The splits are not visible to the user and are managed internally by creating a single entry per split in the database. The problem with that is that each split takes up space in the database and adds overhead every time we load requests. In this PR, I have added support for merging the splits back so that we have a single entry in the database for those requests. The splits would be merged only when we are done processing all the splits for a request.

Special notes for your reviewer:
I have also made the following few changes:

• Changed the order of performing deletion and insertion from a boltdb write batch. Earlier we used to perform insertion before deletion which means if we have duplicate entries lined up in deletion and insertion, we will end up losing the entries since deletion updates were applied at the end. This change was required because we do have some similar database entries between sharded vs merged requests. I have checked the code, and this is the only place where we will have both inserts and deletes in the same batch, so this change should not cause any problems.
• Removed code to parse deletion queries(selector) every time we load requests. Parsing deletion queries every time we load requests is one of the major contributors to increased memory usage. The parsed deletion query is used only when executing deletion, and we have checks in place to parse the query when required in those places.

Checklist

• Tests updated

[paul1r]

Should we log this information before continuing?