Proposal: Add retries to network operations

[gdavison]

In some cases, the Checkpoint or other endpoints may be temporarily unavailable. This can cause callers of, e.g. installer.Ensure(...), to fail intermittently.

For example:

Get "https://releases.hashicorp.com/terraform/1.4.6/index.json": read tcp 192.168.22.118:38854->18.161.6.90:443: read: connection reset by peer

Retrying the network calls could mitigate these failures

[radeksimko]

Hey @gdavison
This is a good idea!

FWIW we have observed same/similar problem popping up sporadically over in terraform-exec E2E nightly tests when running terraform init as providers also rely on releases.hashicorp.com. i.e. if you use hc-install to install Terraform and then execute it, there's more patching to do, also inside Terraform Core.