performance drop due to incorrect logql type parsing

[ftong2020]

Describe the bug

Loki's query parser incorrect parse query

{app="my-app", service="my-service"} |= `some-filter-content` |= ``

as "limit" type instead of filter, which incorrectly set query parallelism to 1, leading to drastic performance drop.

This actually happens a lot since some one browsing logs in grafana and use grafana menu to add filter will easily trigger this. For logs that reaches 100s of gigs, log volume query will finish very fast(with parallel query), but log content itself seems stuck and never returns.

To Reproduce
Steps to reproduce the behavior:

1. Started Loki 3.4.2 in distributed mode
2. Query {app="my-app", service="my-service"} |= some-filter-content |= ``
3. Find logs in query-frontend, confirm it is parsed as limited query

Expected behavior
Query above should be recognized as filter type, thus utilize all available queriers.

Environment:

• Infrastructure: kubernetes
• Deployment tool: helm

Screenshots, Promtail config, or terminal output
There is already a unit test in logql module, I added two lines to demonstrate this issue.