CLN: remove IndexEngine.set_value

[jbrockmendel]

made possible bc Series._values now returns DTA/TDA for datetime64/timedelta64

Small perf improvement

In [3]: dti = pd.date_range('2016-01-01', freq='D', periods=10**4)              
In [4]: idx = list('abcdefghijklmnop')                                          
In [5]: arr = np.random.random(len(idx)*len(dti)).reshape(len(dti), -1)         
In [6]: df = pd.DataFrame(arr, index=dti, columns=idx)                          

In [7]: %timeit df2 = df._set_value("2043-05-14", "c", 4)                      
381 µs ± 14.3 µs per loop (mean ± std. dev. of 7 runs, 1000 loops each)   # <-- master
330 µs ± 13 µs per loop (mean ± std. dev. of 7 runs, 1000 loops each)  # <-- PR

[jreback]

i think much more clear if you reverse these no? (L588 first), then bool casting is out of the way.

[jbrockmendel]

I'm not clear on what you're suggesting here. There is no bool casting in this function.

[jreback]

reverse the order of the if elif, IOW handle the bool dtype.type first

[jreback]

would not be averse to you importing validate_numeric_casting at the top

[jreback]

can you not just do

elsif (.....) as by-definition they are not bools at this point.

[jbrockmendel]

the case where issubclass(dtype.type, np.integer) will pass through in both cases i think

[jreback]

ok this is fine