fix: rewrite handling of missing keys for indexers

[cristianmtr]

This will be required for the CRUD integration tests suite

• handling of missing keys should not Raise
• missing keys are filtered out of the request
• we do log a warning - up to the user to interpret if that's expected or not

[github-actions]

Latency summary

Current PR yields:

• 😶 index QPS at 1898, delta to last 3 avg.: +0%
• 😶 query QPS at 37, delta to last 3 avg.: +0%

Breakdown

Version	Index QPS	Query QPS
current	1898	37
0.9.2	1853	37
0.9.1	1987	37
0.8.22	1880	36

Backed by latency-tracking. Further commits will update this comment.

[github-actions]

Latency summary

Current PR yields:

• 😶 index QPS at 1758, delta to last 3 avg.: +0%
• 😶 query QPS at 33, delta to last 3 avg.: -2%

Breakdown

Version	Index QPS	Query QPS
current	1758	33
0.9.2	1776	33
0.9.1	1723	33
0.8.22	1732	33

Backed by latency-tracking. Further commits will update this comment.

[cristianmtr]

Not sure why I can't reply to this comment: #1628 (comment)
but @JoanFM here's the reply

When we call an .update, the keys will be computed for the update anyway. The update will then call the delete with only the filtered out keys.

If we don't do it this way, the .delete method will have to recalculate and filter out the missing keys.

When .delete is called on its own (not part of an update), then the optimization will not happen.

[maximilianwerk]

Just organizational/cleanup comments.