route metric on multihomed ec2 instances is based on mac address instead of device-number

Bug Description

Since #114 all network interfaces get IPs and a default route, with a metric based on `nic_idx`. `nic_idx` isn't derived from the EC2 device-number, but from sorting by MAC address:

https://github.com/canonical/cloud-init/blob/6600c642af3817fe5e0170cb7b4eeac4be3c60eb/cloudinit/sources/DataSourceEc2.py#L766

A machine with multiple interfaces, where the MAC addresses aren't in order, will end up with the wrong metric. For example, if eth0 has a MAC address that sorts before eth1, the default route for eth1 will be given metric 100, while eth0 will get 200.

This wasn't picked up in tests (e.g. https://paste.ubuntu.com/p/jw6Wws5Jhk/) because interfaces are typically assigned in MAC address order. This isn't true when creating an additional network interface in EC2, or creating an instance from pre-existing interfaces.

This should say "if eth0 has a MAC address that sorts *after* eth1, the default route for eth1 will be given metric 100, while eth0 will get 200"

A pull request for this is open at https://github.com/canonical/cloud-init/pull/342

(That PR landed.)

This bug is believed to be fixed in cloud-init in version 20.3. If this is still a problem for you, please make a comment and set the state back to New

Thank you.

Tracked in Github Issues as https://github.com/canonical/cloud-init/issues/3667