fix: make read_from_internal_position inline

[theVerySharpFlat]

read_from_internal_position is essentially a wrapper for bitset_count_range, so inlining here makes sense.

Interestingly, this improves my buddy_bench time by about half a second!

[spaskalev]

The myth about compilers being smart is busted once again, eh :) I dislike how I had to violate the neat position abstraction in the find free function and go deep with the internal read but that also was a significant performance improvement.

Thanks for looking into it!

[theVerySharpFlat]

The myth about compilers being smart is busted once again

right? This is sort of similar to the previous patch. I expected the compiler (gcc 15, mind you) to inline memset, but here we are. And you would almost expect -O3 to violate user inline hints to improve performance.

Thanks for looking into it!

Of course! Thanks for the quick reply!