Charris pep8 numpy lib for 1.9

Some of those problems look like potential coding errors. In those
cases a Fixme comment was made and the offending code, usually an
unused variable, was commented out.

Replaces the current method to zero items, from multiplication to
using `np.where`.

Replace an explicit type check with setting `copy=False` in call to `astype`.

Speeds calculation up by ~3x for 100x100 matrices, and by ~45x for
1000x1000

reduces buffer copy and comparison overhead for boolean outer product

* `np.tri` now produces less intermediate arrays. Runs about 40% faster for
  general dtypes, up to 3x faster for boolean arrays.
* `np.tril` now does smarter type conversions (thanks Julian!), and together
  with the improvements in `np.tri` now runs about 30% faster. `np.triu`
  runs almost 2x faster than before, but still runs 20% slower than
  `np.tril`, which is an improvement over the 50% difference before.
* `np.triu_indices` and `np.tril_indices` do not call `np.mask_indices`,
  instead they call `np.where` directly on a boolean array created with
  `np.tri`. They now run roughly 2x faster.
* Removed the constraint for the array to be square in calls to
  `np.triu_indices`, `np.tril_indices`, `np.triu_indices_from` and
  `np.tril_indices_from`.

The documentation misrepresented what happened, leaving out division
by the total number of sample points.

Also run spellcheck over function_base.py and twodim_base.py and break
some long lines.

Closes #2423.