Add float32 circuits and 32-bit arithmetic support

Increase float16 randomized test coverage

Run input-coverage sweep by default

Add float16 classification LUT ops

Add float16 deg2rad/rad2deg circuits

Add float16 constant circuits

Add checked outputs for domain ops

Add float16 domain flags

Add degree-mode trig LUTs

Add float16 scientific LUT ops

Rebuild float16 LUT/pow + 16-bit arithmetic; fix neg16bit

Add randomized float16 tests and document open issues

Make inputs coverage strict and document full verbose runs

Add semantic tests for orphan selector/comparator tensors

Add inputs-coverage sweep and document coverage run

Remove dead routing code from eval.py

Fix sbc8bit test: pass borrow input to subtractor eval

Fix broken circuits: bitwise shifts, symmetry8bit, subtractor eval

Expand eval coverage, fix mod2/mod4/mod8 circuits

Fix float16.mul/div circuit bugs and eval tests

Unify eval.py and arithmetic_eval.py into single evaluator

Fix inference functions for mul/div/toint/fromint

Add float16.sub tests (115/115 passing)

float16.add: include guard bit in arithmetic (123/125 tests)

Add float16.add circuit (93/125 tests passing)

Add float16.normalize, neg, abs circuits

Add CLZ16BIT, fix README claims, update TODO

Add CLZ8BIT and float16 circuits (unpack, pack, cmp)