refactor: share the env status read across providers

[ianaya89]

Six status data sources — aws, gcp, azure, hcloud, k8s and aws_hosted — each carried the same ~105 line Read: fetch, nil check, short-circuit on the applied spec revision, build a poll closure, wait, re-fetch, nil check, set state. Only the client method, the response field and the model type differed.

The copies had already drifted:

• "env status matchs spec" in azure, gcp, hcloud and k8s; aws and aws_hosted spelled it correctly
• the // Re-fetch to populate the model comment existed in two of the six

Both are the kind of thing that only shows up as a diff between siblings nobody diffs.

Shape

common.ReadEnvStatus owns the flow. Each provider supplies one refresh closure that fetches the status, writes it into its model and returns what the wait needs (AppliedSpecRevision, Errors, Found). WaitForSpecRevision becomes unexported and takes that same closure, which drops the PollFunc type and stops threading the env name through it.

Per-provider Read goes from ~105 lines to ~45, and what is left is the binding — client method, response field, model — with no control flow.

Also collapses the DISCONNECTED branch in the wait, whose two arms both did blockingErrors = append(...) then continue:

if (e.Code == "DISCONNECTED" || e.Code == "K8S_DISCONNECTED") && result.AppliedSpecRevision == 0 {
    ...
    continue
}
blockingErrors = append(blockingErrors, e)

Behaviour

Unchanged, with one exception worth flagging: the read timeout is now parsed before the short-circuit rather than only on the waiting path, so a malformed timeouts block is reported in both cases instead of only when a wait happens.

Tests

env_status/common had no tests; the flow now has 11, covering no target, target already met, refresh error, not found, refresh-after-wait, timeout, DISCONNECTED/K8S_DISCONNECTED before first provision, disconnect after provisioning, a plain provisioning error, and a blocking error arriving alongside a non-blocking disconnect.