crosscluster/logical: in DLQ entry, surface if conflict occurred on local write or replicated write

Currently, a customer has to go through a [manual process](https://cockroachlabs.atlassian.net/wiki/spaces/CKB/pages/4084465707/Playbook+Logical+Data+Replication#Dead-Letter-Queue-Management) to understand if replicating row ended up in the DLQ due to a conflict with a row also written by LDR or from a foreground workload on the destination.

We should annotate the error in the dlq entry write path (code[ ref)](https://github.com/msbutler/cockroach/blob/master/pkg/crosscluster/logical/dead_letter_queue.go#L198) to indicate if the conflicting row has an originTimestamp (from LDR, rare) or not (foreground workload, expected). We could add a new column to the DLQ table too, but maybe thats more complicated than this needs to be.

For this to work nicely, i think we need to teach constraint violation error handling to surface the conflicting row's origin timestamp.

Jira issue: CRDB-55456

Start at pkg/crosscluster/logical/dead_letter_queue.go around line 198, then trace the constraint-violation error handling that produces the DLQ entry. Determine how to expose the conflicting row's originTimestamp and distinguish replicated from foreground conflicts; done means the DLQ error identifies which kind of write caused the conflict.

Written by the indexing model from the issue text.