Possible duplicates in the decision table

[ColinHebert]

The current logic behind the decision caching system assumes that there should be only one entry per searchee_id/guid pair in the decision table https://github.com/cross-seed/cross-seed/blob/master/src/decide.ts#L215-L224

The database setup though does not accomodate for that: https://github.com/cross-seed/cross-seed/blob/master/src/migrations/00-initialSchema.ts#L12

To ensure true uniqueness of searchee_id/guid in decision, the following should be added to the table:

table.unique(['searchee_id', 'guid'])

This change can't be made without doing some cleanup ahead of time though, as there are likely "duplicate" entries in the existing decision table. So a migration script, which probably should remove all decisions but the last ones in case of duplicates, will be necessary.

---

In addition to the table constraints themselves, the logic of decision creation relies on calls to assessAndSaveResults (not exported) https://github.com/cross-seed/cross-seed/blob/master/src/decide.ts#L185-L203 which may happen multiple times.

A more elegant way of handling it would be for assessCandidateCaching (exported as assessCandidate) to "attempt to create a decision" ahead of time (with dummy values), and make the necessary updates when needed. This would be akin to https://github.com/cross-seed/cross-seed/blob/master/src/pipeline.ts#L75-L79 which attempts to create a searchee ahead of time to ensure a unique Id is available for follow up operations.

Discussion on discord: https://discord.com/channels/880949701845872672/880949702349164556/1162863467129225257