feat: implemented three halves garbling scheme

[themighty1]

This PR adds the three-halves garbling scheme which (compared to the half-gates scheme) improves communication complexity by ~25% at the cost of computational overhead.

Native evaluator shows ~10% slowdown:

half-gates:

evaluator/st/1M         time:   [208.63 ms 208.81 ms 208.99 ms]
                        thrpt:  [4.8080 Melem/s 4.8120 Melem/s 4.8162 Melem/s]

evaluator/mt/1M         time:   [48.138 ms 49.133 ms 49.820 ms]
                        thrpt:  [20.169 Melem/s 20.451 Melem/s 20.873 Melem/s]

three-halves:

evaluator/st/1M         time:   [226.17 ms 226.28 ms 226.40 ms]
                        thrpt:  [4.4382 Melem/s 4.4405 Melem/s 4.4427 Melem/s]
                
evaluator/mt/1M         time:   [52.164 ms 52.895 ms 53.831 ms]
                        thrpt:  [18.666 Melem/s 18.996 Melem/s 19.262 Melem/s]

WASM garbler shows ~30% slowdown:

=== garble ===
Name                                      Median (ms)  Per-iter (us)  AND gates/s
----------------------------------------------------------------------------------
garble/half_gates_garbler_1m                 31159.06      308661.15       3.26M
garble/three_halves_garbler_1m               43586.99      437999.43       2.29M

[sinui0]

This is a very large PR. Can we start with just the core crate?

[themighty1]

superceded by #367