[8.x](backport #42966) fix(libbeat): mitigate race condition in ratelimit processor

[mergify]

Proposed commit message

This PR fixes race conditions on the ratelimit processor. The race conditions can only manifest with configurations that have a global ratelimit processor and multiple inputs. Fixed by adding a mutex to the bucket struct.

During benchmarking, I also noticed an unnecessary allocation occurring in the hot path. Removing this allocation resulted in a 25% performance improvement. While locks have an inherent cost, in this scenario, the performance gain offsets that cost.

Checklist

• My code follows the style guidelines of this project
• I have commented my code, particularly in hard-to-understand areas
• I have made corresponding changes to the documentation
• I have made corresponding change to the default configuration files
• I have added tests that prove my fix is effective or that my feature works
• I have added an entry in CHANGELOG.next.asciidoc or CHANGELOG-developer.next.asciidoc.

How to test this PR locally

Run filebeat with a global ratelimit processor under the race detector:

$ cd x-pack/filebeat; go build -race .
$ GORACE="halt_on_error=1" ./filebeat -c ratelimit.yml -d '*' -v -e

filebeat:
  inputs:
    - type: benchmark
      enabled: true
      message: "bench1"
      eps: 100
    - type: benchmark
      enabled: true
      message: "bench2"
      eps: 100
    - type: benchmark
      enabled: true
      message: "bench3"
      eps: 100
    - type: benchmark
      enabled: true
      message: "bench4"
      eps: 100
processors:
- rate_limit:
   limit: "1/s"
output.console:
  pretty: true
logging:
  level: debug
  selectors:
    - "*"
path.home: /tmp/filebeat

---

This is an automatic backport of pull request #42966 done by [Mergify](https://mergify.com).

[elasticmachine]

Pinging @elastic/elastic-agent-data-plane (Team:Elastic-Agent-Data-Plane)