enable SqlOrchestrationService.PurgeInstanceStateAsync to delete more than 1000 instances

[usemam]

Supposed to fix #297.

Summary of changes

• New PurgeInstanceStateByFilter stored procedure (logic.sql): Accepts @CreatedTimeFrom, @CreatedTimeTo, and @RuntimeStatusFilter parameters, selects all matching instance IDs server-side, and delegates deletion to the existing PurgeInstanceStateByID sproc. This follows the same pattern as the existing PurgeInstanceStateByTime procedure.
• Rewritten PurgeInstanceStateAsync(PurgeInstanceFilter) (SqlOrchestrationService.cs): Replaced the two-step query-then-delete approach (limited to 1000) with a single call to the new stored procedure, removing the artificial limit entirely.
• Integration test (DataRetentionTests.cs): Added PurgesMoreThanOneThousandInstances test that creates 1100 completed orchestrations and verifies all are purged in a single PurgeInstanceStateAsync call.

[cgillum]

Thank you for this PR! One concern I have is that the list of instances to be purged could grow extremely large, in which case this transaction may never be able to succeed due to memory constraints on the database. Should we consider a design where we keep the original implementation (which includes the 1000 PageSize limit) but run it in a loop so that we purge in batches of 1000, thus capping the maximum size of the database transaction?

[usemam]

Sounds good. I was initially apprehensive about iterative approach, but in the case of extremely large pool of instances to be purged I'd agree that it'll perform better.
Pushed my latest changes - please take a look when possible.
Thank you!

[cgillum]

LGTM, thanks!

[cgillum]

I just now triggered the CI to validate the changes.