The program runs remotely isn't line-buffered and ssh is just passing
through what it gets when it gets it.
Refer to: https://unix.stackexchange.com/questions/21920/ssh-output-isnt-line-buffered
This commit add -tt argument to ssh and forces tty allocation.
Now each line will show up in local machine immediately after it shows remotely.

* Previously there is a cpp file called xxx__ALLcls.cpp, which includes
  nearly all other cpp files. Such a big cpp file will be compiled. And
  it can not leverage parallel jobs with Makefile.
* To enbale parallel build, we should pass VM_PARALLEL_BUILDS=1 to the
  Makefile to generate emu.

Use fake Icache to fetch 8 instructions per cycle.

* now mainargs is supported by nexus-am

Otherwise dmemXbar will not behave correctly.

* remove --trace
* split file and functions with smaller unit

* also put mainargs at 0x80001000 to match AM

* now we can pass the image used in AXI4RAM as an argument of emu

* it seems faster than sbt during the start time