Tweak and improve opentf-ctl in CI scenario

Current situation

• The warmup and polling delays are hard-coded (1s, and 5s or 10s depending on the performed queries).
• A dryRun query parameter was added to the POST /workflows endpoint, but we cannot use it from opentf-ctl/
• There are no easy ways to specify additional tags when running a workflow, which would be handy when targeting a specific set of execution environments.
• opentf-ctl get workflows can get slow if there are many workflows. There is a expand=manifest query parameter we could use.
• opentf-ctl get workflow shows execution environment requests every 5s. This can be annoying.

Desired outcome

A new set of orchestrator options, --warmup-delay, --polling-delay, and --max-retry options (integers, in seconds for the first two), that can be specified on command line for commands performing requests.

Those parameters could also be defined for orchestrators in the OPENTFCONFIG file.

A --dry-run option can be specified when running a workflow. It does everything as usual, but does not start the workflow. (Files and required environment variables still must be provided.)

A --tags tag[,tag] option can add tags to the jobs. It should handle replacing the targeted os, if any.

opentf-ctl get workflows should no longer be unbearably slow when showing a few tens of workflows.

Execution environment requests are now only shown once per job in interactive mode. (Using the --verbose (-v) option shows them all.)