support forcast and value increase/reduce too fast(percent) in alert rule

currently, InfluxDB OSS alerting rules are too weak than other enterprise level monitoring tools. It's hard for us to choose influxdb oss cluster as monitoring stack.
1. forcast: disk full in specifical time, http request count
2. increase or reduce too fast: there is no actul or well measured threshold for metrics
3. add percentile support
4. add native or plugin support for ML: like prophet abnormal detection, STL(could be provided in enterprise subscription)

The issue names no files, tests, or entry points. Start by locating the alert-rule subsystem and scoping the requested forecasting, rate-of-change thresholds, percentile support, and ML integrations separately. Done would require a concrete, agreed scope with implementation and tests for the selected capability.

Written by the indexing model from the issue text.