This can happen when Custom Resource Definitions do not exist on the
cluster. For example, because only a subset of the Flux controllers are
installed on the cluster.

Previously, the detection was based on a combination of error type and
string matching. However, a more reliable (and maintained)
`apimeta.IsNoMatchError` checker is available upstream. Making it less
likely this suddenly stops to matching properly when Kubernetes changes
things.

Signed-off-by: Hidde Beydals <hidde@hhh.computer>
(cherry picked from commit bf6754e2)