Investigate memory overflow possibility when reading large collections

[gagik]

Some of the read tools simply run an i.e. aggregate or read query and load everything into memory. This can be quite problematic for a collection with millions of items.

We should test this against a very large collection and come up with mitigations (perhaps impose a hard limit or do a count and prompt the user to give a limit if the size is too large).

[nirinchev]

We should also look into streaming responses so that we don't load everything in memory: https://github.com/cyanheads/model-context-protocol-resources/blob/main/guides/mcp-server-development-guide.md#streaming-responses-transport-feature