Challenge - Reduce float32 Multiplier Latency

The module multiplies finite normalized float32 values or zero. The starter code waits several cycles before returning the result.

Challenge objective

• keep the same interface;
• accept i_valid='1' as the start of a transaction;
• produce o_valid='1' and the correct product in at most 5 cycles;
• handle sign, biased exponent and hidden mantissa;
• handle zero inputs;
• i_reset='1' clears internal state.

The goal is timing improvement: split the computation into shorter registered steps.