Simply bumping TextIOWrapper._CHUNK_SIZE from 128 to 1024 yields some
significant performance improvements.

./python -m timeit -s "f=open('LICENSE')" "f.seek(0)" "for line in f: pass"
 -> with 128: 6.21 msec per loop
 -> with 1024: 3.41 msec per loop

./python -m timeit -s "f=open('LICENSE')" "f.seek(0)" "while
f.read(100): pass"
 -> with 128: 4.03 msec per loop
 -> with 1024: 1.25 msec per loop

Is there any counter-indication to doing so?