Add support for flen=64

[AntonLydike]

Currently still missing a lot of the D extension (all except fmadd, fmsub, fnmsub, fnmadd, fadd, fsub, fmul, fdiv, fsqrt, fsgnj, fsgnjn, fsgnjx, fmin, fmax, feq, flt, fle, fld, fsd), missing conversion and move instructions.

This may break some of the float32 stuff, so we should be very careful with this.

Todos:

• Running with --flen=32 breaks because it tries to load the D extension but can't
• Write changelog
• Maybe refactor out some misc bugfixes in this

[AntonLydike]

@superlopuh any instructions you are missing so far?