Support for unaligned memory

[recp]

Currently all mat4 and vec4 operations need memory to aligned 16 because of SSE/AVX instructions, this library may support unaligned memory operations, this would beuseful when using library on non-library type e.g. float[4][4]:

for instance: needs aligned:

glm_mat4_mul(  )

no aligment required:

glm_mat4u_mul(  )