Fix delete snapshot so that it doesn't orphan data keys or delete the wrong key

[vasil-pashov]

Reference Issues/PRs

Fixes: #1958
Fixes: #1863

What does this implement or fix?

• Fix orphaned keys. recurse_index_keys does not assume that the all
  atom keys are for the same stream id anymore
• Fix AggregationClause::process vector iterator invalidation. Not
  related to the issue but C++ were crashing at runtime because of this.
  Calling erase invalidates the iterator in a loop.

Any other comments?

Checklist

Checklist for code changes...

• Have you updated the relevant docstrings, documentation and copyright notice?
• Is this contribution tested against all ArcticDB's features?
• Do all exceptions introduced raise appropriate error messages?
• Are API changes highlighted in the PR description?
• Is the PR labelled as enhancement or bug so it appears in autogenerated release notes?

[poodlewars]

Change looks correct to me, could you share the results of repeating the benchmark from the AtomKeyPacked PR?

[vasil-pashov]

Results of running time -v for the new change

       Prune took 1.9612538814544678s

	Command being timed: "python -m pytest /home/vasil/Documents/source/ArcticDB/python/tests/integration/arcticdb/version_store/test_snapshot.py::test_prune_previous_memory_usage"
	User time (seconds): 46.35
	System time (seconds): 10.89
	Percent of CPU this job got: 82%
	Elapsed (wall clock) time (h:mm:ss or m:ss): 1:09.31
	Average shared text size (kbytes): 0
	Average unshared data size (kbytes): 0
	Average stack size (kbytes): 0
	Average total size (kbytes): 0
	Maximum resident set size (kbytes): 1375964
	Average resident set size (kbytes): 0
	Major (requiring I/O) page faults: 1
	Minor (reclaiming a frame) page faults: 4926581
	Voluntary context switches: 228693
	Involuntary context switches: 57025
	Swaps: 0
	File system inputs: 0
	File system outputs: 3706960
	Socket messages sent: 0
	Socket messages received: 0
	Signals delivered: 0
	Page size (bytes): 4096
	Exit status: 0