Optimize memory procedures

[Fumuran]

What should be done?

Memory procedures, which load only one element from the memory, should use dedicated element-oriented procedures instead of word-oriented (mem_load instead of mem_loadw). This will decrease the number of cycles required for element obtaining.

How should it be done?

Code of the procedures which load just one memory element should be updated to use element-oriented procedures.

When is this task done?

This task is done when every memory procedure will be optimized.

Additional context

See also: #1126 (comment)