WarpXSolverVec: Remove duplicates

[WeiqunZhang]

This appears necessary for PETSc to work.

[ax3l]

🎉

[JustinRayAngus]

Below is a comparison of the running-average numbers of GMRES iterations/Newton iteration from a 2D planar pinch problem with the duplicates (left) and without the duplicates (right). The number of GMRES iterations is significantly improved with the duplicates removed, permitting a practical value for the additive Schwarz method (asm) overlap value.

The setup for this test is a 224x16 grid with 100 ppc for each species at t = 0. The initial time step is such that wpe0*dt = 36 with wpe0 the plasma frequency measure with the initial density (ne = ni = 1e17/cm^3). These simulations were performed using 56 Dane cores (8x8 boxes). The figure on the left (with duplicates) is obtained using PR #6396 with a mass matrix pc width = 1. The figure on the right (without duplicates) is obtained with a branch where PR #6396 is merged with this PR.

[JustinRayAngus]

Thanks!