Multi-RHS support for Prolongator and Restrictor

[maddyscientist]

This PR adds support for multi-RHS to both the prolongator and restrictor. This will be an essential building block for multi-RHS multigrid solvers:

• Multi-RHS support for prolongator and restrictor (supports up to 64 RHS)
  • Prolongator sees a <6x speedup with 24 RHS (on Volta)
  • Restrictor sees a <2.5x speedup with 24 RHS (on Volta)
• Update the orthonormality test in MG::verify to use multi-RHS
• Cleanup to MG class to move from heap allocation (pointers) to stack allocation (objects) for the ColorSpinorField objects
• Optimization to the fine-grid prolongator / restrictor workflow: fine-grid basis change is now handled in line by the transfer operators
• Fix a long-standing performance bug where aliasing field copies were not elided
• Other minor miscellaneous improvements

[weinbe2]

This passes a visual review as the code stands. Upon a more detailed review of the generated code, there's some stack spillage at staggered-size Nc --- @maddyscientist 's is going to look into it, and if it's easy to fix will do, if not we'll get this merged in regardless and I'll file an issue to revisit it.

[weinbe2]

Compile:

cmake -DCMAKE_BUILD_TYPE=RELEASE -DQUDA_DIRAC_DEFAULT_OFF=ON -DQUDA_DIRAC_STAGGERED=ON   -DQUDA_GPU_ARCH=sm_80 -DQUDA_DOWNLOAD_USQCD=ON -DQUDA_QIO=ON -DQUDA_QMP=ON   -DQUDA_MULTIGRID=ON -DQUDA_MULTIGRID_NVEC_LIST="24,64,96" ../quda

Generate a well behaved 16^4 field:

mpirun -np 1 ./heatbath_test --dim 16 16 16 16 --save-gauge l16t16b7p0   --heatbath-beta 7.0 --heatbath-coldstart true --heatbath-num-steps 10 --heatbath-warmup-steps 1000

Run a 3 <-> 64 <-> 96 test, which flexes recursion:

mpirun -np 1 ./staggered_invert_test \
  --prec double --prec-sloppy single --prec-null half --prec-precondition half \
  --mass 0.01 --recon 13 --recon-sloppy 9 --recon-precondition 9 \
  --dim 16 16 16 16 --gridsize 1 1 1 1 --load-gauge l16t16b7p0 \
  --dslash-type asqtad --compute-fat-long true --tadpole-coeff 0.905160183 --tol 1e-10 \
  --verbosity verbose --solve-type direct --solution-type mat --inv-type gcr \
  --inv-multigrid true --mg-levels 4 --mg-coarse-solve-type 0 direct --mg-staggered-coarsen-type kd-optimized \
  --mg-block-size 0 1 1 1 1 --mg-nvec 0 3 \
  --mg-block-size 1 4 4 4 4 --mg-nvec 1 64 \
  --mg-block-size 2 2 2 2 2 --mg-nvec 2 96 \
  --mg-setup-tol 1 1e-5 --mg-setup-tol 2 1e-5 --mg-setup-inv 1 cgnr --mg-setup-inv 2 cgnr \
  --nsrc 1 --niter 25 \
  --mg-setup-use-mma 0 true --mg-setup-use-mma 1 true --mg-setup-use-mma 2 true --mg-setup-use-mma 3 true \
  --mg-smoother 0 ca-gcr --mg-smoother-solve-type 0 direct  --mg-nu-pre 0 0 --mg-nu-post 0 4 \
  --mg-smoother 1 ca-gcr --mg-smoother-solve-type 1 direct --mg-nu-pre 1 0 --mg-nu-post 1 4 \
  --mg-smoother 2 ca-gcr --mg-smoother-solve-type 2 direct-pc  --mg-nu-pre 2 0 --mg-nu-post 2 4 \
  --mg-coarse-solver 1 gcr --mg-coarse-solve-type 1 direct --mg-coarse-solver-tol 1 0.25 --mg-coarse-solver-maxiter 1 16 \
  --mg-coarse-solver 2 gcr --mg-coarse-solve-type 2 direct-pc --mg-coarse-solver-tol 2 0.25 --mg-coarse-solver-maxiter 2 16 \
  --mg-coarse-solver 3 ca-gcr --mg-coarse-solve-type 3 direct-pc --mg-coarse-solver-tol 3 0.25 --mg-coarse-solver-maxiter 3 16 \
  --mg-verbosity 0 verbose --mg-verbosity 1 verbose --mg-verbosity 2 verbose --mg-verbosity 3 verbose

To test the recursion in the multi-rhs coarse operator, append these lines to perform block TRLM with a block size of 48, which should properly split into {32, 16}:

  --mg-eig 3 true --mg-eig-type 3 blktrlm --mg-eig-use-dagger 3 false --mg-eig-use-normop 3 true \
  --mg-nvec 3 48 --mg-eig-n-ev 3 96 --mg-eig-n-kr 3 192 --mg-eig-tol 3 1e-4 --mg-eig-use-poly-acc 3 false \
  --mg-eig-block-size 3 48 --mg-eig-spectrum 3 SR \
  --mg-eig-max-restarts 3 1000