Alert Threshold to better account for high spikes and low troughs

### Problem Statement

Right now the alerts thresholds are minimal and may miss some alerting spikes and timeframes of low error reporting. For example, if an e-commerce site has a period of high errors one week (think Black Friday) followed by a week of low errors (but still need to be alerted on), or if there is a week, or month, where there are relatively low number of alerts followed by a period of time where alerts are firing far more often than normal.

Is there a way that Sentry can help make the instances of these spikes and troughs more normalized. Ideally, using ML/AI might help make sure that alerts fire when appropriate without either ending up silenced or too noisy when anomalies happen in a given time frame.

### Solution Brainstorm

_No response_

### Product Area

Alerts