[dv/mem_bkdr_util] Improve memory write performance

[matutem]

Description

A vcs time profile for chip_sw_pwrmgr_smoketest shows 50% of the time is spent in uvm_hdl_deposit for mem_bkdr_util::write. This is excessive, and we should be able to reduce it significantly.

I think we should explore alternative implementations, perhaps including direct updates of the memory arrays as in $readmemh, to avoid uvm_hdl_deposit.

[andreaskurth]

Triaged for pwrmgr. Assigning https://github.com/lowRISC/opentitan/labels/Type%3AIcebox as not critical for M2.5.

[matutem]

This performance optimization is not indispensable, and should not get in the way of milestones.