Do not grow memory without bound

[drj11]

extract.py seems to grow without bound when doing its stuff (observe with htop).

Not only is this bad form generally, but on free (which is limited to 512MB) it causes the process to be abruptly killed mysteriously. (which leaves lots of files behind: issue #31)

[pwaller]

Good luck with that.

For a 21mb all_tables.xls it has a maximum resident set of 1.8GB.

[pwaller]

This one is still an issue since writes aren't streaming yet.

[pwaller]

I've just tried xlsxwriter. In the {'constant_memory': True} mode it uses virtually no memory but a lot of CPU time. On my laptop around 14 seconds for ~100krow, compared to "almost nothing" for CSV. I suggest we put in a hard limit of 100krow anyway and use pyexcelerate since it is considerably faster. Currently testing the performance of these in a few scenarios, PR incoming soon.

@drj11 @morty