Merge pull request #19 from Kontinuation/fix-parsing-wkb

Fix read_uint32 and read_uint64 for big-endian machines, potentially improving performance

Reload unaligned doubles

For non arm64/amd64 systems. Minor optimization that loads
unaligned doubles into a seperate array and use that array
directly, rather than use a branch that is intended for swapping
endianness.

Fix compile error on m32 arch

rename buf object

Update README.md

Print the smallest floating point representation.

This commmit makes a change where floating points are now
printed as their smallest textual representation. Before, all
numbers were always ddd.fffff. But now it may be printed with an
exponent, if and when that will make it smaller. This affects wkt
and geojson outputs. The reason for this change is to save space,
both network and disk, but without losing any accuracy.

For example, the number 5000000 will be now be 5e6.

To disable this feature call tg_env_set_print_fixed_floats(true);

Update reference counting

Use a relaxed/release/acquire pattern, and start counter at one.

Update API docs

Initialize memory to silence lto warning in gcc.

Fix gcc warnings