Add state retrieval for alert rules

[alexander-akhmetov]

Updates the existing alert tools to retrieve state information (firing, normal, pending, etc.) for alerts.

Since the Alerting provisioning API does not expose this information, I had to add a custom alerting client that uses the same API used by the UI. This API is not part of the officially stable public APIs, but the client makes it possible to add other endpoints, for example alert state history.

This API supports pagination of the alert rule groups, but I haven't added anything for it yet. Can do a follow-up :)

[alexander-akhmetov]

Noticed this was done in another PR: #81

[sd2k]

This LGTM! I am curious if any of the changes in #95 have affected the use of base URL here - sorry if so!

[sd2k]

This probably doesn't need exporting - ideally we'd just have the actual tools exported from this tools package. (we may have missed some already but I'll try and fix that soon 😅)

(same goes for the various structs below!)

[alexander-akhmetov]

Makes sense! Un-exported :)

[alexander-akhmetov]

I am curious if any of the changes in #95 have affected the use of base URL here - sorry if so!

Nope, all good!

[alexander-akhmetov]

Does it make sense to have something similar to IncidentClient in mcpgrafana.go, AlertingClientFromContext? Or is it fine to initialize the client as it is now since it's the same as Grafana client basically?

[sd2k]

Hm I think we have that separate IncidentClient in the context to enable connection reuse in the stdio case, but it doesn't feel like a big enough deal really (and annoyingly we need to remember to add the client to the ContextFunc for every transport, which is easy to forget). So I'm happy to keep it where it is (I may even stop storing the IncidentClient in context one day).

[sd2k]

LGTM, thanks again! 🥳