ENH: Improve perf of str_extract

[unutbu]

Building the result from a list of strings or a list of lists seems to improve performance over building the result via arr.apply(f).

[unutbu]

./test_perf.sh -b master -t str-extract -r extract

Invoked with :
--ncalls: 3
--repeats: 3


-------------------------------------------------------------------------------
Test name                                    | head[ms] | base[ms] |  ratio   |
-------------------------------------------------------------------------------
strings_extract                              |  39.7120 | 578.9154 |   0.0686 |
-------------------------------------------------------------------------------
Test name                                    | head[ms] | base[ms] |  ratio   |
-------------------------------------------------------------------------------

[jreback]

pls add a note in release notes (u can ref this PR)
in improvements sections
looks good!

[unutbu]

m.groups()[0] might return ''. Only None should be turned into NaN. I'll fix this when I get back.

[ghost]

Always glad to see 10x. tx.

[jreback]

@unutbu ready for merge?

[unutbu]

@jreback: Yes, ready.

[jreback]

thanks!