Feature Request: Multiple Source Alerts For Inhibition Rules

Currently, the alert manager inhibition rules support logic that allows the inhibition of an alert when a different alert exists.
While this functionality is great, it is lacking when more complex gating is desired.

The alert manager does not have a functionality supporting inhibiting an alert based on multiple sources.
For example, the following logic can't be expressed as an inhibition rule:
"Mute all alerts with {alertname="no_info", application="enricher"} if an alert with {alertname="instance_down", application="parser"}' exists AND an alert with {alertname="instance_down", application="writer}" exists."

This becomes increasingly apparent when wanting to define alerting rules which represent systems that don't follow a single, linear flow.

Suggested Solution:

Add support for multiple alert sources in the inhibition rules configuration.

Make the source section of the rule a list of sources, with a logical AND functionality between them. For example:

inhibit_rules:
  - source:
     - matchers:
        - alertname="instance_down"
        - application="parser"
       equal: ["cluster"]
     - matchers:
        - alertname="instance_down"
        - application="writer"
       equal: ["severity"]
    target_matchers:
      - alertname="no_info"
      - application="enricher"

Notes:

1. Supporting only logical AND is sufficient, since the alert manager already supports a logical OR, just by splitting the expression into two separate rules.

2. the equal field is configured per source matchers set. This is important to avoid coupling between sets.
   Returning to the above example, the following alerts:
   {alertname="instance_down", application="parser", cluster="A", severity="3"}
{alertname="instance_down", application="writer", cluster="B", severity="2"}
   will cause the following alert to be inhibited:
   {alertname="no_info", application="enricher", cluster="A", severity="2"}
   Note that the first source alert's severity does not match the target alert's severity, and that the second source alert's cluster does not match the target alert's cluster.

3. While this change can seems insignificant, it opens a door for a whole new world of advanced inhibition rules that were not possible to implement previously. Therefore, although this can result in some changes to the configuration, this feature (or one with a similar functionality) could be well worth the effort.