Transaction is being rolled back automatically by some process on Ignite 2.15 and 2.16

Hi community,

Recently when trying to migrate a tool from Ignite 2.14 to 2.15 or 2.16, I started to see that our puts were being rollback automatically by some internal process on Ignite, specifically we are getting the following exception:

`org.apache.ignite.internal.transactions.IgniteTxRollbackCheckedException: Failed to finish transaction because it has been rolled back [timeout=60000, tx=GridDhtTxLocal[xid=2e9746e0091-00000000-13a
6-dc0d-0000-000000000001,......`

Please see [ignite_trasaction_rollback-log.txt](https://github.com/user-attachments/files/15812031/ignite_trasaction_rollback-log.txt) for a complete stack trace.

This is my configuration:
1. Number of threads: 10 (putting records into caches).
2. Transaction Concurrency: PESSIMISTIC
3. Transaction Isolation: REPEATABLE_READ
4. Transaction timeout: 60 seconds
5. Cache Atomicity Mode: TRANSACTIONAL

My exact same code works on prior to these new 2 versions, so something new was added that now it is causing a problem about transactions. Could you please help to determine what could be the reason of the problem?

Thanks in advance and regards,

Start by reviewing the attached ignite_trasaction_rollback-log.txt and comparing the reported behavior between Ignite 2.14, 2.15, and 2.16. Reproduce the rollback with the listed pessimistic, repeatable-read transaction settings and determine the regression or internal process responsible; done means identifying a confirmed cause and a targeted fix or workaround.

Written by the indexing model from the issue text.