Avoid uint64 for 32bit compatibility

Rather than worry about truncation casting from a possibly 64bit value
down to a possibly 32bit size_t we just limit the total bytes per
invocation to 4G using an unsigned integer.

Thanks to @seriyps for the report.

Fixes davisp#61

Disable strict aliasing

This is motivated by issue davisp#60 to see if we can get away with `-O3`
while still handling the cast of a `Decoder**` to `void**`.

Fixes davisp#60

Include documentation on the maps feature

Fix timeouts for short doubles

Timeouts apparently don't actually work unless you use a test generator.
Which makes sense in hindsight.

Expand the README to cover the API more clearly

Yield back to Erlang while encoding JSON

This adds a configurable limit on the number of bytes produced by
the encoder before yielding back to the Erlang VM. This is to avoid the
infamous scheduler collapse issues.

The `jiffy:encode/2` now takes an option `{bytes_per_iter,
pos_integer()}` that controls the yield frequency. The default value is
2048.

Noncharacters U+FFFF and U+FFFE are not invalid.

Properly decode them and accept to encode them.

This was clarified by Unicode Technical Committee:
http://www.unicode.org/versions/corrigendum9.html

Add check for below array bounds

Signed-off-by: Nobuhiro Iwamatsu <[email protected]>

Prevent segfaults on unterminated strings

A single quote input was causing segfaults due to sneaking past the
string termination logic. This patch corrects that lapse in conditional
by only parsing strings where a closing quote was found. All other
strings are rejected as invalid.

Big thanks to Jean-Charles Campagne (@jccampagne) for reporting the
issue.

Minor change for building on Windows