Make the hashing function of SMT terms faster

[aqjune-aws]

Previously, the hash functions of a few datastructures in SMT were first generating their repr string, then hashed it. Since this is slow, this pull request adds their simpler implementations.

According to some internal benchmarks, with --no-solve, the total amount of time for file writing was significantly reduced.

Previous numbers:

    Solver/file writing                      17039       370.4        4912    43.4%
  VC discharge                               24826       539.7        7212    63.2%
SMT verification                             32391       704.2        8197    82.5%
-------------------------------------------------------------------------------------
TOTAL                                        39276       853.8        8484   100.0%

After this optimization:

    Solver/file writing                       6461       140.5        1783    22.8%
  VC discharge                               14128       307.1        4363    49.8%
SMT verification                             21681       471.3        5236    76.4%
-------------------------------------------------------------------------------------
TOTAL                                        28378       616.9        5357   100.0%

Also checked that this did not change the number of verified/unverified programs.

By submitting this pull request, I confirm that you can use, modify, copy, and redistribute this contribution, under the terms of your choice.

[aqjune-aws]

Thanks, the patch is much simpler than before.