Enhance the performance of the frontend JSON codec

[damnever]

What this PR does:

Stream json encoding using existing functions, reducing CPU usage ~50%.

Which issue(s) this PR fixes:

After upgrading from v1.16 to v1.19, the frontend's CPU usage increased significantly. The primary cause appears to be related to JSON codec, the use of encoding/json and the read-and-parse process. This patch addresses both issues. However, despite applying the patch, CPU usage remains higher than it was prior to the upgrade. The exact reason for this is unclear, maybe it is due to Go 1.24 or something else, but profiling results still show that decoding labels continues to consume a substantial amount of processing time (profile below).

Checklist

• Tests updated
• Documentation added
• CHANGELOG.md updated - the order of entries should be [CHANGE], [FEATURE], [ENHANCEMENT], [BUGFIX]

[yeya24]

Thanks. I think this change makes sense!
Maybe @alanprot can also take a look

[yeya24]

I am wondering if we have a better package to import? In case we remove some code from that chunk package.

[alanprot]

maybe we can just something like this?

var _ = labels.Labels{}

[damnever]

maybe we can reimplement the codec for cortexpb.LabelAdapter and use it directly

[yeya24]

I am worried that at sometime we just remove that code path from the package without noticing and we have the regression.

We don't have to do it in this PR but let's at least add some TODO or notes for re-implementing the codec or something similar. Creating an issue also works

[damnever]

We can export the legacy functions directly and use them without relying on global registration.

[damnever]

da7343d

[alanprot]

Thanks!

[alanprot]

maybe we can just something like this?

var _ = labels.Labels{}

[alanprot]

Just out of curiosity, can you also show the impact in memory on this?

[damnever]

@alanprot

[yeya24]

@damnever For the graph you shared, when was the change deployed and how much it impacted memory?

[damnever]

@yeya24 the middle of 06/13 12:00 - 06/14 00:00, peak memory decreased significantly, possibly by 70%

[danielblando]

@damnever

I think this PR broke the integration tests.
This is constantly failing on my PR after the rebase

    querier_test.go:534: 
        	Error Trace:	/home/runner/work/cortex/cortex/integration/querier_test.go:534
        	Error:      	Received unexpected error:
        	            	unable to find metrics [thanos_store_index_cache_requests_total] with expected values. Last error: <nil>. Last values: [31]
        	Test:       	TestQuerierWithBlocksStorageRunningInSingleBinaryMode/blocks_sharding_enabled,_redis_index_cache,_bucket_index_enabled,thanosEngine=false

https://github.com/cortexproject/cortex/actions/runs/15765121553/job/44441732594?pr=6827

Looks like it also failed in your tasks, but the metrics is different
https://github.com/cortexproject/cortex/actions/runs/15729526842/job/44327445878