AWSMachinePool subnet filters ignored if no matches found

[AverageMarcus]

/kind bug

What steps did you take and what happened:

Create an AWSMachinePool CR with a subnet filter that doesn't match any existing subnet in the account.
E.g.

subnet:
  filters:
    - name: tag:subnet-role
      values: ["non-existent"]

When no subnets are found to match the provided filter the SubnetIDs function in the autoscaling service will use an empty array for subnetIDs (rather than populating it with the IDs of any subnets found to match the filter). This empty array is then passed on to the SubnetIDs function on the scope which mistakenly treats the empty array of subnets IDs as no desired subnets were provided on the AWSMachinePool CR. This will then fall back to fetching any subnet that matches the provided availability zone (or the subnets of the AWSCluster if no AZs provided).

What did you expect to happen:

If no matching subnet is found to match the provided filters an error should be shown instead to avoid mistakenly placing instances into incorrect subnets.

Anything else you would like to add:
[Miscellaneous information that will assist in solving the issue.]

• More discussion can be found in this Slack thread
• AWSMachine CRs respect the filter and will display an error if no matching subnet is found.
• Proposed solution: have the SubnetIDs function of the autoscaling service return an error if inputFilters are provided but subnetIDs is empty at the end of the function.

Environment:

• Cluster-api-provider-aws version: >= 1.5
• Kubernetes version: (use kubectl version):
• OS (e.g. from /etc/os-release):

[Skarlso]

/triage accepted

[AverageMarcus]

/assign