Parallel tc to ttgt

[pthomadakis]

This PR introduces a parallel implementation of TC-to-TTGT pass and parallel implementation for OptDenseTransposePass

As an example, the performance of the following intensli.ta kernel:

def main() {
	#IndexLabel Declarations
	IndexLabel [a, b, d] = [1024];
	IndexLabel [c] = [512];

	#Tensor Declarations
	Tensor<double> v([d, c, a], {Dense});
	Tensor<double> t2([b, d], {Dense});
	Tensor<double> i0([a, b, c], {Dense});

	#Tensor Fill Operation
	v[d, c, a] = 2.3;
	t2[b, d] = 3.4;
	i0[a, b, c] = 0.0;


        var time0 = getTime();
	i0[a, b, c] = v[d, c, a] * t2[b, d];
        var time1 = getTime();
	printElapsedTime(time0, time1);
        var out = SUM(i0[a,b,c]);
        print(out);
}

Improves ~6x on my 8-core CPU (~24s to ~4s).
However, we still need to investigate why we are behind numpy and why cometpy is way behind