ASV: used integer ndarray as indexer for Series.take indexing benchmark

[hardikpnsp]

• closes BENCH: Index.take benchmark is measuring wrong thing #18000
• tests added / passed
• passes black pandas
• passes git diff upstream/master -u -- "*.py" | flake8 --diff

Series.take was using a list of booleans as indexers. Now, It is replaced with a random list of integers.
Here are the benchmark results with asv run --bench indexing.Take

Benchmarks Results:

Before:

[100.00%] ··· indexing.Take.time_take                                                                                                                                                ok
[100.00%] ··· ========== =============
                index                 
              ---------- -------------
                 int      4.13±0.08ms 
               datetime   4.13±0.01ms 
              ========== =============

After:

[100.00%] ··· indexing.Take.time_take                                                                                                                                                ok
[100.00%] ··· ========== =============
                index                 
              ---------- -------------
                 int      6.36±0.08ms 
               datetime   6.29±0.03ms 
              ========== =============

[jorisvandenbossche]

Thanks for working on this!

[hardikpnsp]

Updated benchmarks after changing indexer from list to ndarray:

[100.00%] ··· indexing.Take.time_take                                                                                                                                                ok
[100.00%] ··· ========== =========
                index             
              ---------- ---------
                 int      336±2μs 
               datetime   348±2μs 
              ========== =========

[jreback]

thanks @hardikpnsp