CR2-22 CR2-49 CR2-48 add metrics list to sdk

[robghchen]

Adding list custom metrics functionality to sdk and cli for jobs, experiments, deployments, and notebooks

QA Test Plan:

1. in cli, check that this new list command shows up in the help text: gradient deployments metrics --help
2. create a deployment (gui or cli doesn't matter) and wait for it to finish provisioning
3. try the new list command: gradient deployments metrics list --id XYZ, should return custom metrics (a list of words),
4. repeat 1 and 2 for jobs, experiments, and notebooks. these might return null but that's okay for now, just checking that this new command is available. note the stuff that returns null below if any and I'll solve that in a separate ticket.

[linear]

CR-22 View custom metrics from CLI

• Allow querying custom metrics using experiment metrics or experiment stream. Right now we use a hardcoded list of options and we would like to relax the validation requirement to allow any input. We should update the help for these commands to list all of the known available options (found here).
• We need a new metrics sub-command for each workload, list, which will query available custom metrics from the metrics server.

# all three of these go to the metrics service on the cluster
# talks to api to get metrics url, then talks to metrics url
gradient experiments metrics list      # optional timeframe, done over https
# --> names of metrics
gradient experiments metrics stream    # query, done over ws, long-lived
gradient experiments metrics get       # query + timeframe, done over https
# --> actual metric values

[robghchen]

this can be ready for review, i made a separate PR for messing around with tests #331

[colin-welch]

This might be too broad of a check. How about localhost?

[robghchen]

at first i assumed it would contain localhost but when i logged out metrics_api_url it said something along the lines of http...local.paperspace...

[robghchen]

oh i could check local.paperspace instead duh lol

[colin-welch]

If this is specific to that domain then maybe we should be including the port in the metrics URL returned for workloads in that domain, ie. in the service layer.

[robghchen]

yea that would be better

[tallyw00d]

list exists. Deployments, experiments, jobs, notebooks all currently return null

LGTM

[PSBOT]

🎉 This PR is included in version 1.2.3 🎉

The release is available on GitHub release

Your semantic-release bot 📦🚀