Add an endpoint to bulk request metric history for experiment runs

[jonburdo]

Is your feature request related to a problem? Please describe.

Currently, fetching the metric history for multiple experiment runs requires making a separate API call for each run.

Describe the solution you'd like

Add a new batch endpoint that can return the metric history for a specific metric across multiple specified runs in a single API call.

The suggested endpoint is:
GET /experiment_runs/metric_history?experimentRunIds=[RUN_ID_1,RUN_ID_2,...]&metricName=[METRIC_NAME]

Describe alternatives you've considered
The obvious alternative is to continue using the existing /experiment_runs/{experimentrunId}/metric_history endpoint and have the client make repeated requests.

Additional context

This is related to work for mprahl/kubeflow-community#1

[dhirajsb]

An alternative to experimentRunIds query param is to leverage the filterQuery swiss army knife. We should add support for it for both the experiment and experiment_run metric history endpoints to allow users to filter metrics by name, customProperties of the metrics, etc.