GS/HW: Reduce sw/hdr colclip in more cases.

[lightningterror]

Description of Changes

GS/HW: Reduce sw/hdr colclip in more cases.
When doing Cs*Alpha + Cd*(1 - Alpha) or Cd*Alpha + Cs*(1 - Alpha) with an alpha of 128 or lower we don't really need to hdr or sw colclip blend it because colour range of the result will be between 0-1 (0-255) without it overflowing.

Also update previous optimizations to include Ad cases when RTA is already scaled.

Rationale behind Changes

More speed, less barriers/draw calls/copies.

Suggested Testing Steps

Notable dump mentions:

dbz_bt2_oi_imine_depth Draw Calls: -26 [991=>965] Render Passes: -25 [69=>44] Barriers: -1 [12=>11] Copies: -25 [27=>2]
Power_Rangers_-_Super_Legends_SLUS-21679_20221215160520 Draw Calls: -20 [109=>89] Render Passes: -26 [56=>30] Barriers: -7 [9=>2] Copies: -20 [21=>1]
Tales_of_Destiny_Directors_Cut_SLPS-25842_20221015223409 Render Passes: -1 [7=>6] Barriers: -29 [29=>0]

Benchmark/test the following dumps on vk/gl/dx:
Dumps.zip

List of affected games/dumps to test:
GameList.txt

[JordanTheToaster]

Some benchs that mark of 2 games that showed the most improvement.

[kamfretoz]

Some benchmarks of the dumps on AMD GPU: