The validation start/end times generation is broken on other OSs than Linux

[Nuttymoon]

Currently, we use the pipe lookup plugin to dynamically generate the validation start/end times to add validators to Subnets:

ansible-avalanche-collection/roles/subnet/defaults/main.yml

Lines 72 to 77 in d53a31c

	subnet_validator_start_time_command_default: 'date -d "2 minutes" --rfc-3339=seconds | tr " " T' # in 2 minutes
	subnet_validator_end_time_command_default: 'date -d "1 week 2 minutes" --rfc-3339=seconds | tr " " T' # in 1 week and 2 minutes
	subnet_validator_start_time_command_darwin: "date -v+2M '+%Y-%m-%dT%H:%M:%SZ'" # in 2 minutes
	subnet_validator_end_time_command_darwin: "date -v+1w -v+2M '+%Y-%m-%dT%H:%M:%SZ'" # in 1 week and 2 minutes
	subnet_validator_start_time: '{{ lookup("pipe", lookup("vars", "subnet_validator_start_time_command_"+ansible_os_family | lower, default=subnet_validator_start_time_command_default)) }}'
	subnet_validator_end_time: '{{ lookup("pipe", lookup("vars", "subnet_validator_end_time_command_"+ansible_os_family | lower, default=subnet_validator_end_time_command_default)) }}'

This won't work on MacOS because the os_family is actually evaluated on the Ansible target, thus taking the debian value. The lookup will then fail silently, and we end up with empty strings.

A better way to generate the start/end times would be to execute the date command on the Ansible targets and register a fact there.