Read graph tool should be more efficient instead of pulling every entity in the database

[thesobercoder]

Guidelines

Please note that GitHub issues are only meant for bug reports/feature requests.
Before sending unsolicited pull requests, please open a feature request to discuss your proposal.

If you have questions on usage, please ask on the Discussion Forum instead of creating an issue here.

Feature description (Mandatory)

Currently, the read graph tool returns every possible entity saved in the database. Although this works, it is not an efficient strategy as this will result in too many token usages.

Considered alternatives

I think like search nodes, the read graph should also perform full text searches to begin with on the search index to limit the number of results.

How this feature can improve the project?

This will greatly improve the token usage and not pollute the LLM context with unnecessary context.