String broadcastables

[kwmsmith]

This starts the process of expanding Blaze's string column support to include upper and lower. This is useful since having first-class (and optimized) support for common string operations is useful for the string-munging pain points that users hit.

I'm marking these as part of the "experimental" API currently, since I'm not wild about the str_upper, str_lower, etc. naming scheme. I'd like to find a better naming system for these if we can.

We have immediate need for upper and lower, so I'm putting these in for 0.10.

Regarding naming schemes: I like the Pandas' style df.col.str.upper().str.replace(...). We could expand that to include df.col.dt.datetimemethod() for datetimes as well.

We'll have to think about which string and datetime methods we want to support, for which backends, and what are the semantics when the method in question returns multiple values. All of that is outside the scope for this PR.

This PR also deprecates strlen and adds str_len for consistency.

[kwmsmith]

ping @ltransom

[necaris]

FWIW 👍 for the Pandas style of df.col.str.*, especially because that lets us do more with df.col.dt.*, df.col.int.*, and so on.

[llllllllll]

I am also +1 on the str and dt descriptor namespaces.

[sandhujasmine]

Tested on postgres test data in blaze-benchmarks. No issues.

[sandhujasmine]

Created an epic to capture this feature but I'm not able to create the epic and add issue to it. Might be a permissions issue.

Here's the high level epic: #1476
Here's one issue for the fist bullet point: #1475