Make testing with JIT options easier and more precise

[mgaudet]

So, to see how the patch is holding up against trunk I experimented with rebasing on top of a0acd82 and running tests.

What I found is that running with

LD_LIBRARY_PATH=$PWD OMR_JIT_OPTIONS=-Xjit:count=0 make test

I saw huge slowdowns before we even got to executing the actual test files while running the extension makes. I'm not sure what's going on there, but every time you type 'make test' it's regenerating a bunch of things related to the exts and causing a multi-minute delay before make test even gets started.

[mgaudet]

It seems that trunk is rebuilding the exts much more aggressively than ruby_2_4_omr, and when building exts, there's a series of calls to miniruby like this:

./miniruby -I./lib -I. -I.ext/common  ./ext/extmk.rb --make='make' \
	--command-output=ext/json/exts.mk --dest-dir="" --extout=".ext" --mflags="" --make-flags="" --gnumake=yes --extflags=""  --extstatic  \
	-- configure ext/json

each of which pays all the startup overhead of a count=0 compilation, dramatically slowing the get-to-test process.

The fix here is to make passing JIT options for testing Ruby dramatically more precise.