fix(deployments): stop full scans and coalesce cache misses

[romeroyonatan]

Issue

1. Deployment range queries scanned every GitHub page resulting in slow deployments dashboards
2. Concurrent cache misses duplicated requests:

panel A: cache miss → mutex unlock → GitHub request
panel B: cache miss → mutex unlock → GitHub request

Solution

1. Paginate until deployments predate the requested range
2. Coalesce identical deployment cache misses.

Implementation detail

singleflight.Group tracks in-flight work by cache key, so identical misses share one request. The field deploymentRequests singleflight.Group is deployment-specific because this PR intentionally changes only deployments

[cla-assistant]

All committers have signed the CLA.

[cla-assistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}

[perrito666]

wouldn't createdAt.After(from) && createdAT.Before(to) be more idiomatic? also bear in mind that this is < not <= just in case

[romeroyonatan]

For this PR, no: the requirement is inclusive [from, to].

!createdAt.Before(from) && !createdAt.After(to)

means:

createdAt >= from && createdAt <= to

Using:

createdAt.After(from) && createdAt.Before(to)

would mean:

createdAt > from && createdAt < to

and would incorrectly exclude deployments exactly at from or to.

The current form is idiomatic for inclusive time.Time bounds because Go has no
BetweenInclusive helper.

This pattern is also present in other code in this repo. For example:

https://github.com/grafana/github-datasource/blob/main/pkg/github/workflows.go#L102