The human-centric platform for production ML & AI

Outerbounds

We have been running into default cards failing to render for tasks that have `pandas.DataFrame` artifacts which have multi-indices with `pandas.Timestamp` levels. It looks like <https://github.com/Netflix/metaflow/pull/1344|#1344> improved parsing of dataframe columns, and also datetime indices, but it did not address multiindexes or (single-level) indexes of other types. Here's a PR to use the same logic that's used to parse columns for indexes. <https://github.com/Netflix/metaflow/pull/1574>

As an aside, the datetime format string `time_format = "%Y-%m-%dT%H:%M:%SZ"` doesn't handle timezones correctly. It should be `%Z` at the end; as is it is equivalent to assuming all datetimes are UTC. Here is another PR to fix that. <https://github.com/Netflix/metaflow/pull/1575/files>