Support Group By in Metric Monitors for Spans (title edited)

### Problem Statement

Feedback from Discord:

> Somewhat related I have created an alert for p95(transaction.duration) > 300ms. Based on my tests it looks like the p95 is computed across all transactions and not after grouping by transaction.name. Is that correct? If so how can I compute the p95 after grouping? Such a grouping is important because say I have 2 endpoints myApp/XXX and myApp/YYY and 99% of the calls go to myApp/XXX which always returns in 100ms I will not be able to detect if any the myApp/YYY calls take an exceptionally long amount of time.
>
> Setting the alerts individually will not work as we have 100s of different transaction names

### Solution Brainstorm

_No response_

### Product Area

Alerts