RI-HFX forces

[abussy]

Implementation of RI-HFX forces based on sparse DBCSR tensor contractions. The new code is an extension of the work by @pseewald. New features cover:

• RI-HFX forces in the RHO and MO flavors
• Possibility of using a shorter range RI metric (for both flavors)
• Stress tensor

Additionally, the option of using RI-HFX was generalized throughout the code. It is now available for (although only in the RHO flavor):

• RPA
• MP2 (including gradients)
• TDDFPT (not including forces)
• LINRES

Some of the SCF RI-HFX routines where refactored for performance. In particular, alloc_block_3c now avoids allocating tons of arrays and build_3c_integrals was OMP parallelized.

Relevant regtests were added.