Draft: Laplacian metaGGA

This MR adds support for Laplacian level MetaGGA functionals. I currently track here the progress in the verification of the results and code cleanups.

Some notes on the changes introduced so far:

• fft_laplacian_g2r has been added to speedup computation of the Laplacian of the charge density. There are many other ways (see lines after 249 in the new version of PW/src/v_of_rho.f90). One option should be adopted and the rest of the code removed before merging.
• Absolute value has been added on line 346 (original version) / 409 (new version) of libxc. Not sure why it was not there in the first place.
• Laplacian contribution is only computed when needed: no impact on gKS MetaGGA simulations is expected. On the contrary, the component depending on tau of the KS potential is always computed, even when tau is zero.
• Removed warning in XClib/dft_setting_routines.f90
• Important: OR has been replaced with AND in lines 320 / 430 and 337 / 450 of XClib/xc_wrapper_mgga.f90. The condition tau=0 removed the XC potential contribution for atoms sitting exactly at the origin of the Cartesian coordinate system.

Verification

Abinit already implements Laplacian level metaGGA so a direct comparison of the absolute value of the total energy is possible, thanks to the pseudopotential conversion script available here.

All simulations have been performed with libXC version 6.2.2 compiled with -DDISABLE_FHC=ON (see here). This is important for performing SCAN simulations as discussed here. All results reported below use r2SCANL (input_dft = 'XC-000I-000I-000L-000L-718L-719L') with ONCV_PBE pseudopotentials in upf format converted to psp8 format (NB core_correction=F for all pseudo used). The cutoff is set to 80 Ry.

Lattice	QE tot. E (eV)	Abinit tot E (eV)	QE - Abinit (meV)
BCC Li (a=6 a.u.)	−194.265 543 8	−194.273 434 2	7.89
FCC Cu (a=6.6 a.u.)*	−4.978 935 250 × 103	−4.978 932 910 × 103	-2.34

* Cu does not converge to 1e-9 Ry for both QE and Abinit.