iaas: stackit_server panics when GetServer fails during update

[devpie]

Description

stackit_server crashes the provider process when the GetServer call at the start of Update
fails. The error is logged and execution continues with a server that is nil on that path, and a
few lines further down that value is dereferenced.

stackit/internal/services/iaas/server/resource.go, main @ a7d114b:

// :825
if server, err = r.client.DefaultAPI.GetServer(ctx, projectId, region, serverId).Execute(); err != nil {
	core.LogAndAddError(ctx, &resp.Diagnostics, "Error retrieving server state", fmt.Sprintf("Getting server state: %v", err))
}   // no return

// :840
if err := updateServerStatus(ctx, r.client.DefaultAPI, server.Status, &model, region); err != nil {

Any failing GetServer reaches this — a 500, a timeout, a transient network error. The operator
gets a plugin crash instead of the "Error retrieving server state" diagnostic that was just
recorded.

The identical missing return sits at :577 in the create path. There it is caught by accident:
mapFields rejects a nil response at :949 and the caller returns at :583-585. The visible
effect is only a misleading diagnostic — Processing API payload: response input is nil instead of
the call that actually failed.

Steps to reproduce

The configuration is unremarkable; any stackit_server that is updated while the API is briefly
unavailable will do:

resource "stackit_server" "example" {
  project_id   = var.project_id
  name         = "example"
  machine_type = "g1.1"
  boot_volume = {
    size        = 32
    source_type = "image"
    source_id   = var.image_id
  }
  network_interfaces = [var.nic_id]
}

1. terraform apply to create the server.
2. Change name (or any attribute handled by Update).
3. terraform apply again while GET /v2/projects/{p}/regions/{r}/servers/{id} answers 500.

The create path can be driven deterministically with a mock server; a test for it is in the PR
below and fails without the fix.

Actual behavior

The provider panics with a nil pointer dereference at resource.go:840 and Terraform reports that
the plugin crashed, discarding the diagnostic the provider had already added.

Expected behavior

Update returns after adding the diagnostic, so the operator sees "Error retrieving server state"
and the reason.

Environment

• OS: macOS 15 (darwin/arm64)
• Terraform version: not version-specific
• Version of the STACKIT Terraform provider: main @ a7d114b

Additional information

Found with NilAway while surveying the provider for the class
of defect in #1743. resource.go:840 is reported on main and gone with the fix. It is a distinct
defect from #1743 — that one is about responses the SDK decodes to nil, this one is a missing
return. Fix in the PR linked below.