internally int1 is int8

1. compiler knows all the registers
2. but only 64bits register is a color (for reg alloc)
3. backend records the length of GPR for each operand during instruction
selection
4. after reg alloc, when replacing temp with a color, find corresponding
GPR for the length recorded before

currently we only use 64bit registers, extend/truncate result if it is
not 64bits. This is awful. Going to use all the registers, for example,
AL, AX, EAX, RAX, but will not use AH (so reg alloc is simpler)

currently only translate one old test into new one with macros,
the macro set is going to grow as it needs to. The macros are
located under tests. atm I do not intend to expose it for clients.

test NEW instruction, also test thread transformation and worked around libmu library loading problem

add array sum test; add linked list reversal test; add simple test for new and thread transformation; refactor to allow compiling through C backend

Note: made two places in api implementation as unimplemented as
the instruction/type chnaged.

going to test it with test_jit suite