Optional BT_USE_DOUBLE_PRECISION

[not-fl3]

For now, bullet is compiling with #define BT_USE_DOUBLE_PRECISION, resulting f64 as floating point format everywhere. It will be useful to make that define as a build option.

Here is two problems needed to fix first:

• Ensure that SharedMemory is working without BT_USE_DOUBLE_PRECISION and not using hardcoded double (while AFAIK it use and a lot)
• Make bindgen as a build step - to generate f32/f64 version.
• Make tests for both variants (with a macro or something)