Optimize correct colordict entries

[damusss]

as @Starbuck5 noticed string methods are called even if the color name existed raw in the color dictionary. This PR adds a fast path when that is the case. I tested with timeit before and after this PR and correct names got a major performance increase (from 5.5 to 1.8) while other cases are basically left unchanged. I believe a fast path should also be considered for hex colors, but it's beyond the scope of this PR. MRE:

import pygame
import timeit

# change the string for different cases
print(timeit.timeit("pygame.Color('white')", globals=globals(), number=10_000_000))

# OLD
# correct: 5.5-5.8
# hex: 8-8.2
# case: 5.6-5.8
# space: 5.9+

# NEW
# correct: 1.8-1.9
# hex: 8-8.2
# case: 5.7-5.9
# space: 6+

[ankith26]

LGTM, thanks for the PR 🎉

[Starbuck5]

Nice work, you turned me complaining about something into something productive! I did not have this in mind at all.

I tested with set_at to get an example of it doing something real (but also cheap).

import time
import pygame

pygame.init()

screen = pygame.Surface((300,300))

start = time.time()
for _ in range(1000000):
    screen.set_at((4,35), "red")
print(time.time() - start)

This went from 0.67 seconds to 0.29 seconds, a more than 2x boost.

Saving 0.38 seconds over a million reps means this saves 380 nanoseconds per call. If we assume the mapped tuple is absolutely zero color mapping cost there are only 120 nanoseconds left to save.

For the curious, using a mapped color was 0.17 seconds, using a 3 element tuple took 0.25 seconds, using a 4 element tuple took 0.26 seconds, and using a hex string took 1.1 seconds.