Fixing reduction operations

[Joao-Pedro-Cabral]

Problems in the current implementation:

• Synchronization bugs with the following instructions (register pack operation ends much later compared to pipeline wrapper).
• Many extra cycles are spent shifting the result element within the result buffer.

Solution:

• Write the result element directly to the lowest bits of the result buffer (no shift)

[ParkerJones567]

This issue might still be present in vicuna2.0, please test with the current development branch