Update abandon tx functionality to drop related attempts

[amit-momin]

BCFR-1078

Context

The abandon transaction functionality marks unstarted, in_progress, and unconfirmed transactions as fatal_error and clears out any assigned nonces. However, attempts for in_progress or unconfirmed transactions remained in the db even though these attempts aren't needed for any other TXM process. This was causing a conflict with purged transactions that are marked as fatal with attempts that still need to be bumped and confirmed. The only difference is that purge transactions maintain their nonces

Changes

• The abandon transaction functionality was updated to drop attempts when transactions are initially marked as fatal and the nonce is removed.
• The query to find attempts that require a receipt fetch was updated to include a nonce null check.

[dimriou]

I think a bigger underlying design problem exists with how we treat purgeable attempts. Here's what will happen if we have multiple in-flight transactions, including a purgeable attempt. Purgeable attempts are classified as fatal but still treated as in-flight. If we have two transactions with nonce 10 and 11 (the first one is unconfirmed, the second one is purgeable) if we hit Abandon then only the first one will be marked as fatal and the second one won't be impacted. When the TXM restarts, it will try to sync the nonce, look at the DB and see the nonce of the purgeable attempt. Assuming the tx with nonce 10 never made it on-chain this will effectively create a nonce-gap.

[huangzhen1997]

lgtm