Add --version

Created by: scottstanfield

I apologize for adding three specific feature requests to a single issue, but since they're all related, and there's a corresponding screen shot, I thought it might be best here.

Also, I like the idea of the tool emitting Markdown compatible files, but since it's new and unexpected behavior, I would advocate that one needs a flag to explicitly enable it.

Screenshot

Top screen jasmine is running an older version of csvlook. By inspecting the file /usr/local/bin/version I think this is csvkit 0.9.1. Fails when trying to use unknown flag ``--no-inference` instead of ignoring the flag.

Bottom screen belle doesn't have a version, and it's not clear from inspecting the python source. But it's recent as of yesterday.

The CSV file in question is 511 rows x 546 cols (a mix of strings, ints, floats and booleans)

My requests

Add --version. It'll be hard to report bugs if there's not an easy way to query what version is installed. Also, if there was a reliable --version flag, then we could query at shell login and set appropriate wrappers to enable `
Make --no-inference the default state. So one would use --inference to enable it.
Inference is slow. In the second screenshot, it takes 10 seconds to complete, vs. 1 second disabled. The older version (in jasmine) was even faster at 1/2 second.

I love the tool...I probably use csvlook 50-100 times per day. Across 6 machines (I do a lot with CSV files).

My hack solution

Fork csvkit and disable inference by default. If I get this working, I'll submit pull request, but it would be a surprise to your new users that expect the inference to be enabled by default.

Maybe there's an environment variable that can control all this?