[optimizer] cache the previous hash in TransformCtx

[mgree]

#30806 adds some metrics on transforms, but the work of hashing is non-trivial on very large queries. We can reduce hashing by half: store the previous hash in TransformCtx. (Using TransformCtx is @ggevay's idea; I was going to do something much more invasive.)

This shows a ~3s speedup on the InsertMultiRow benchmark on my machine (8s -> 5s).

Motivation

• This PR fixes a recognized bug.
  https://github.com/MaterializeInc/database-issues/issues/8832

Checklist

• This PR has adequate test coverage / QA involvement has been duly considered. (trigger-ci for additional test/nightly runs)
• This PR has an associated up-to-date design doc, is a design doc (template), or is sufficiently small to not require a design.
• If this PR evolves an existing $T ⇔ Proto$T mapping (possibly in a backwards-incompatible way), then it is tagged with a T-proto label.
• If this PR will require changes to cloud orchestration or tests, there is a companion cloud PR to account for those changes that is tagged with the release-blocker label (example).
• If this PR includes major user-facing behavior changes, I have pinged the relevant PM to schedule a changelog post.

[ggevay]

Looks good, thanks!

[mgree]

Running locally, here's the overhead with the soft assert:

NAME                                | TYPE            |      THIS       |      OTHER      |  UNIT  | THRESHOLD  |  Regression?  | 'THIS' is
--------------------------------------------------------------------------------------------------------------------------------------------------------
InsertMultiRow                      | wallclock       |           1.876 |           0.793 |   s    |    10%     |    !!YES!!    | worse:   2.4 TIMES slower
InsertMultiRow                      | memory_mz       |        2126.694 |        2031.326 |   MB   |    20%     |      no       | worse:   4.7% more
InsertMultiRow                      | memory_clusterd |         219.536 |         293.541 |   MB   |    50%     |      no       | better: 25.2% less

NAME                                | TYPE            |      THIS       |      OTHER      |  UNIT  | THRESHOLD  |  Regression?  | 'THIS' is
--------------------------------------------------------------------------------------------------------------------------------------------------------
InsertMultiRow                      | wallclock       |           1.990 |           0.924 |   s    |    10%     |    !!YES!!    | worse:   2.2 TIMES slower
InsertMultiRow                      | memory_mz       |        2055.168 |        1873.016 |   MB   |    20%     |      no       | worse:   9.7% more
InsertMultiRow                      | memory_clusterd |         225.735 |         219.345 |   MB   |    50%     |      no       | worse:   2.9% more

NAME                                | TYPE            |      THIS       |      OTHER      |  UNIT  | THRESHOLD  |  Regression?  | 'THIS' is
--------------------------------------------------------------------------------------------------------------------------------------------------------
InsertMultiRow                      | wallclock       |           2.036 |           0.820 |   s    |    10%     |    !!YES!!    | worse:   2.5 TIMES slower
InsertMultiRow                      | memory_mz       |        1979.828 |        1870.155 |   MB   |    20%     |      no       | worse:   5.9% more
InsertMultiRow                      | memory_clusterd |         220.203 |         257.206 |   MB   |    50%     |      no       | better: 14.4% less

and without it (because I'm not sure whether or not the soft assert is enabled during these feature benchmarks; by the numbers, it seems to not be):

NAME                                | TYPE            |      THIS       |      OTHER      |  UNIT  | THRESHOLD  |  Regression?  | 'THIS' is
--------------------------------------------------------------------------------------------------------------------------------------------------------
InsertMultiRow                      | wallclock       |           1.962 |           0.887 |   s    |    10%     |    !!YES!!    | worse:   2.2 TIMES slower
InsertMultiRow                      | memory_mz       |         805.187 |         984.192 |   MB   |    20%     |      no       | better: 18.2% less
InsertMultiRow                      | memory_clusterd |         195.503 |         229.168 |   MB   |    50%     |      no       | better: 14.7% less

NAME                                | TYPE            |      THIS       |      OTHER      |  UNIT  | THRESHOLD  |  Regression?  | 'THIS' is
--------------------------------------------------------------------------------------------------------------------------------------------------------
InsertMultiRow                      | wallclock       |           1.966 |           0.880 |   s    |    10%     |    !!YES!!    | worse:   2.2 TIMES slower
InsertMultiRow                      | memory_mz       |         894.070 |        1035.690 |   MB   |    20%     |      no       | better: 13.7% less
InsertMultiRow                      | memory_clusterd |         225.163 |         224.113 |   MB   |    50%     |      no       | worse:   0.5% more

NAME                                | TYPE            |      THIS       |      OTHER      |  UNIT  | THRESHOLD  |  Regression?  | 'THIS' is
--------------------------------------------------------------------------------------------------------------------------------------------------------
InsertMultiRow                      | wallclock       |           1.966 |           0.894 |   s    |    10%     |    !!YES!!    | worse:   2.2 TIMES slower
InsertMultiRow                      | memory_mz       |         963.211 |         969.887 |   MB   |    20%     |      no       | better:  0.7% less
InsertMultiRow                      | memory_clusterd |         183.105 |         217.438 |   MB   |    50%     |      no       | better: 15.8% less

So there's definitely still a slowdown, but not quite as egregious as it was before.