Matches-index (July 2025) -- reviving the left-pad index

Because there are very different problems that happen to use floats as their solution. Physics engines (and scientific simulations) typically need to carefully control speed vs precision trade-offs, and sometimes need very very specific ways of calculating expressions, so that objects don't fall through the floor. Similarly ray tracers must be fast, but also have carefully bounded errors in some calculations, since they deal with values that differ by many orders of magnitude and could suffer from systemic errors that add up. There's audio processing, where most of the time there's plenty of precision headroom, but performance must be predictable, and subnormal floats can be a gotcha. There's pixel graphics, where often precision doesn't matter, there's just lots of basic arithmetic that needs to be fast. Compilers being careful about edge cases like NaN or rounding of 52nd bit when blending a couple of 8-bit values is just slowing things down. There's machine learning, UI, animations, that just need some fractional values. There are microcontrollers that just need to convert a few sensor readings.