Potentially pinning an epoch for long time in long transaction

[wenym1]

Describe the bug

Currently in our batch query, for a single query, we will unpin the epoch very soon when all iterators of hummock have been created. However, in a long transaction, after we start the transaction, we will not unpin the epoch before the user manually stop the transaction, which may cause the epoch to be pinned for a long time. Pinning an epoch for long time will make the safe epoch of hummock lags behind the max committed epoch for a long time, and the data size will be large because we need to preserve the multiple versions of a key in our sst.

Therefore we need some mechanism to handle this corner case in case that a accidentally started transaction affect our the performance of our system.

Error message/log

No response

To Reproduce

No response

Expected behavior

No response

How did you deploy RisingWave?

No response

The version of RisingWave

No response

Additional context

No response

[wenym1]

Resolved by #14566