GLES3: Don't call glTexParameter* for invalid filter and repeat modes

[LRFLEW]

This fixes #79315. That issue was caused by TextureStorage::_clear_render_target calling Texture::gl_set_filter and Texture::gl_set_repeat with the value *_MAX. The gl_set_filter and gl_set_repeat functions weren't designed to be called with the values (which appear to be used to force the next calls to make glTexParameter* calls), which results in whatever happens to be the currently bound OpenGL texture to be assigned unexpected parameters. This PR updates Texture::gl_set_filter and Texture::gl_set_repeat so that they only make calls to glTexParameter* when p_filter/p_repeat is a valid value.

I also made some small changes to these functions that shouldn't affect functionality, but figured would be nice to cleanup while I'm here. I removed the (IMO) unhelpful comments and updated how max_lod is assigned to be more consistent with the other local variables in gl_set_filter. I also changed the default values of Texture::state_filter and Texture::state_repeat to better represent their uninitialized state. This shouldn't have an effect on the engine, as these values get initialized in TextureStorage::_texture_set_data either way.

The only part of this I'm not sure about is the handling of *_DEFAULT as the argument. From what I can tell, they should never be used with these functions, so I didn't add a specific case for them, and chose to handle them the same a *_MAX (or any invalid enum value). If this needs to be changed, let me know what behavior you would want these to have, and I can update this PR.

This PR targets the master branch, but it should be safe to cherry-pick for a 4.1.X release. This PR may interact with #79568 and may require a rebase if that gets merged first.

[clayjohn]

Looks great! _DEFAULT should definitely be handled like _MAX neither is technically a valid input to this function, but _MAX is handy to invalidate the cache. In either case avoid gl calls makes sense.

[LRFLEW]

I rebased off the current master branch to resolve the merge conflict

[YuriSizov]

Thanks!

[YuriSizov]

Cherry-picked for 4.1.2.