Fix memory accounting leak in aggregation

[arhimondr]

Description

All the memory pool related failures from #11275 I checked are INSERT queries with a memory leak reported in either an AggregationOperator or an HashAggregationOperator.

I managed to reproduce it locally only once by running the INSERT INTO invalid_partition_value VALUES (4, 'test' || chr(13)) query in a loop and checking whether cluster memory pools are free upon competition.

The issue is incredibly difficult to reproduce as there are two things that must go wrong:

1. A TableWriterOperator instance must happen to get used after close (fixed by Ensure no methods are called on operator after close)
2. The TableWriterOperator operator must trigger an internal aggregation state update to cause it's memory context to get updated

Under normal circumstances the OperatorContext is destroyed during an operator close what prevents any further memory allocations for a given operator. However the OperatorContext for a nested operator within the TableWriterOperator wasn't being properly closed (fixed by Destroy OperatorContext for nested operators).

Also i found that memory can be still be reserved after memory context close (by using tryReserve). Fixed by Ensure no memory can be allocated after memory context close

Is this change a fix, improvement, new feature, refactoring, or other?

Fix

Is this a change to the core query engine, a connector, client library, or the SPI interfaces? (be specific)

Core

How would you describe this change to a non-technical end user or system administrator?

Under rare circumstances DML queries (INSERT,CTAS, etc.) could not release reserved memory what could potentially result in cluster eventually running out of memory.

Related issues, pull requests, and links

Fixes #11275

Documentation

(x) No documentation is needed.
( ) Sufficient documentation is included in this PR.
( ) Documentation PR is available with #prnumber.
( ) Documentation issue #issuenumber is filed, and can be handled later.

Release notes

(x) No release notes entries required.
( ) Release notes entries required with the following suggested text:

# Section
* Fix some things. ({issue}`issuenumber`)

[findepi]

Failure may be related

Error:  Failures: 
Error:    TestDriver.testBrokenOperatorAddSource:300 
Expecting a cause with type:
  <"io.trino.spi.TrinoException">
and message:
  <"Driver was interrupted">
but type was:
  <"java.lang.RuntimeException">
and message was:
  <"Interrupted">.

[losipiuk]

LGTM % CI failure

[arhimondr]

@findepi @losipiuk I had to add two more commits (Unify Driver#process method,Make sure Driver interrupt always handled properly) to fix the test failure. Please take a look.

[losipiuk]

nit: This is always called with 1; maybe replace with processOnce()?

[arhimondr]

I was thinking about that. It feels like processOnce is less intuitive. Though I don't have a strong opinion here.

[arhimondr]

CI: #12385