Vertex Loader Microoptimization

[K0bin]

Just a couple of tiny things that showed up in the profiler in ended up making a measurable difference.

My test case is the Mario Galaxy comet observatory after loading a save game.
Dual core disabled, CPU EFB access disabled, Zen 3 5900X
Before: 118 FPS
After: 128 FPS

Mario Galaxy is particularly heavy when it comes to primitives, so this probably doesn't make a difference in most other games.

[Pokechu22]

From https://dolphin.ci/#/builders/30/builds/1388/steps/8/logs/stdio:

Dolphin>>> 25:53:355 Core/Boot/Boot.cpp:623 N[BOOT]: Booting DFF: /Users/administrator/dff/ssbm-pointsize_1.dff
Dolphin>>> 25:53:437 VideoCommon/VertexManagerBase.cpp:191 E[Video]: Warning: An error occurred.
Dolphin>>> 
Dolphin>>> VertexManager: Too few remaining index values (9 > 8). 32-bit indices or primitive breaking needed.
Dolphin>>> 
Dolphin>>>   Condition: count <= remaining_index_generator_indices
Dolphin>>>   File: /Users/administrator/buildbot-worker/osx_m1/pr-fifoci-mvk-osx-m1/build/Source/Core/VideoCommon/VertexManagerBase.cpp
Dolphin>>>   Line: 191
Dolphin>>>   Function: PrepareForAdditionalData
Dolphin>>> 
Dolphin>>> Ignore and continue?
Dolphin>>> Dolphin [refs/pull/11673/head] 5.0-19058 | JITARM64 SC | Vulkan | HLE
FIFO log playback failed for /Users/administrator/dff/ssbm-pointsize_1.dff

[K0bin]

I fixed that assert the CI ran into as well as the other review comments.

The PR should be ready for another review.