[transactions] Ignore conflicts when inserting

[joel-u410]

If you ever want to force the indexer to re-index some past transactions, and manually update last_processed_block in the crawler_state where name = 'transactions' ... without the on_conflict_do_nothing you can put the indexer into an infinite loop.

Example:

First, update the crawler_state to "rewind history":

namada-indexer=# UPDATE crawler_state set last_processed_block = 38 where name = 'transactions';
UPDATE 1

Now, run transactions:

2024-12-06T18:02:07.412603Z  INFO transactions: Queried block successfully wrapper_txs=1 inner_txs=1 block=39
2024-12-06T18:02:07.423771Z ERROR shared::error: Database error reason=Failed to insert wrapper transactions in db

Caused by:
    duplicate key value violates unique constraint "wrapper_transactions_pkey"
2024-12-06T18:02:10.686714Z  INFO transactions: Queried block successfully wrapper_txs=1 inner_txs=1 block=39
2024-12-06T18:02:10.690671Z ERROR shared::error: Database error reason=Failed to insert wrapper transactions in db

Caused by:
    duplicate key value violates unique constraint "wrapper_transactions_pkey"
2024-12-06T18:02:14.555980Z  INFO transactions: Queried block successfully wrapper_txs=1 inner_txs=1 block=39
2024-12-06T18:02:14.564664Z ERROR shared::error: Database error reason=Failed to insert wrapper transactions in db

Caused by:
    duplicate key value violates unique constraint "wrapper_transactions_pkey"

... forever looping, never recovering ...

In this case, I'd rather it simply ignore the constraint violation (since already-indexed transactions should never change on-chain, we should not UPDATE them) and move on without failing.

I think this scenario can also occur if you were running two instances of the transactions crawler simultaneously (which you might want to do for HA purposes).

Originally posted by @joel-u410 in #81 (comment)

[mateuszjasiuk]

Right! I'm just curious if we do not potentially hide real error in some cases 🤔