feat(metrics): Add before_send_metric callback

[szokeasaurusrex]

Run metrics through ClientOptions::before_send_metric after scope/default attribute enrichment and before enqueue so applications can filter or mutate metrics before batching.

Stacked on #1063.

Co-authored-by: Joris Bayer joris.bayer@sentry.io

Closes #1025
Closes RUST-170

[linear-code]

RUST-170 Add before_send_metric callback processing in sentry-core

[github-actions]

	Messages
📖	Do not forget to update Sentry-docs with your feature once the pull request gets approved.

Generated by 🚫 dangerJS against 7eb91c1

[lcian]

Did we intend to call this just Attribute (could either be a new type or just a type alias)? I don't remember what was the outcome of that discussion (if any).

[szokeasaurusrex]

Yeah, I think it makes sense to make that change. Likely we should do it as a type alias (rename LogAttribute to Attribute, and make LogAttribute an alias to Attribute). However, I believe we can also do this later, should be a backwards and forwards-compatible change as far as I am aware

[sentry]

Bug: The before_send_metric callback is invoked even when options.enable_metrics is false, because the check happens after the callback has already executed.
_{Severity: MEDIUM}

Suggested Fix

Add an early return guard at the beginning of the capture_metric function to check if self.options.enable_metrics is false. This will prevent prepare_metric and the before_send_metric callback from being called when metrics are disabled, aligning its behavior with the capture_log implementation.

Prompt for AI Agent

Review the code at the location below. A potential bug has been identified by an AI
agent. Verify if this is a real issue. If it is, propose a fix; if not, explain why it's
not valid.

Location: sentry-core/src/client.rs#L585-L587

Potential issue: The `capture_metric` function calls `prepare_metric` without first
checking if metrics are enabled via `options.enable_metrics`. The `prepare_metric`
function then invokes the `before_send_metric` callback. The metric is only discarded
afterward because the `metrics_batcher` is `None`. This leads to unexpected side effects
from the callback executing even when the feature is disabled. This behavior is
inconsistent with `capture_log`, which correctly uses an early guard to prevent
processing when logs are disabled.

_{Did we get this right? 👍 / 👎 to inform future reviews.}