Enhanced bytes limiter with data type param

[justinjung04]

• I added CHANGELOG entry for this change.
• Change is not relevant to the end user.

Changes

Introducing two main changes to enhance bytes-based limiting:

1. Changed BytesLimiter.Reserve(num uint64) err to BytesLimiter.ReserveWithType(num uint64, dataType storeDataType) err such that the bytes limiter logic to be more flexible. This is useful given that processing of postings, series and chunks have different impact to pod resource usage. For example, user can have a bytes limiter that has different weights for different data types.
2. nit: Created queryStats.add func to update query stats info more consistently

Verification

Added unit test and e2e test.

[yeya24]

LGTM. Just a small nit on the changelog

[harry671003]

Great work Justin!

[yeya24]

Maybe @fpetkovski @MichaHoffmann or @saswatamcode can help take a look? Thanks

[yeya24]

I will merge this change tomorrow if there is no objection or new review comments. Since this change is not user facing at all, we can address in future PRs if you have any other comments

[saswatamcode]

Thanks, generally LGTM! I think it can be put in changelog as your Store error will now tell you the type of data that exceeded bytes limit?

Also any plans to expose these to user somehow?

[justinjung04]

Thanks for taking a look @saswatamcode !

I think it can be put in changelog as your Store error will now tell you the type of data that exceeded bytes limit?

I haven't changed the error message returned to users, as they already contained the data type information. The main goal of this change was to enhance the limiter logic with access to the data type, which can be hidden from the users.

Also any plans to expose these to user somehow?

Yes! We have been noticing that resource consumption (CPU and memory) for postings, series and chunks are quite different. For example, we were able to fetch up to 1GB/s of chunks until the CPU is exhausted, while we were only able to fetch 258MB/s of series. We should definitely update the limiter logic to assign different weights per data type.