Code patch by Roger Serwy.

This is a behavior change: before this leading and trailing spaces were
stripped from ASCII parts, now they are preserved.  Without this fix we didn't
parse the examples in the RFC correctly, so I think breaking backward
compatibility here is justified.

Patch by Ralf Schlatterbeck.

Use a threshold of 20 ms instead of 10 ms.

Initial patch by Michael Driscoll, I added the example.

Initial patch by Michael Driscoll, I added the example.

but this is more readable.

Includes exposing a doctype handler from expat through pyexpat.

the code are unnecessary.

Obsolete 2.x method.

Thanks to Nick Coghlan for pointing out that I'd forgotten about class
decorators.

in the simple path (not correctly rounded but faster).

Underflow to zero hasn't changed: _mpd_qexp() internally uses MIN_EMIN,
so the result would also underflow to zero for all emin > MIN_EMIN.

In case digits are left, the informal argument is as follows: Underflow can
occur only once in the last multiplication of the power stage (in the Horner
stage Underflow provably cannot occur, and if Underflow occurred twice in
the power stage, the result would underflow to zero on the second occasion).

Since there is no double rounding during Underflow, the effective work
precision is now 1 <= result->digits < prec. It can be shown by a somewhat
tedious argument that abs(result - e**x) < ulp(result, result->digits).

Therefore the correct rounding loop now uses ulp(result, result->digits)
to generate the bounds for e**x in case of Underflow.