[WIP][SPARK-7360][PYTHON] disable useMemo in Pickler

[mengxr]

@nchammas Just found that we don't need to patch Pyrolite because useMemo is exposed in the constructor in Pyrolite 4.4. So the changes in the PR should be sufficient to see the erformance difference caused by useMemo.

[AmplabJenkins]

Merged build triggered.

[AmplabJenkins]

Merged build started.

[SparkQA]

Test build #32431 has started for PR 6065 at commit c8e69c1.

[SparkQA]

Test build #32431 has finished for PR 6065 at commit c8e69c1.

• This patch fails Spark unit tests.
• This patch merges cleanly.
• This patch adds no public classes.

[AmplabJenkins]

Merged build finished. Test FAILed.

[AmplabJenkins]

Test FAILed.
Refer to this link for build results (access rights to CI server needed):
https://amplab.cs.berkeley.edu/jenkins/job/SparkPullRequestBuilder/32431/
Test FAILed.

[AmplabJenkins]

Merged build triggered.

[AmplabJenkins]

Merged build started.

[SparkQA]

Test build #32546 has started for PR 6065 at commit c721a23.

[SparkQA]

Test build #32546 has finished for PR 6065 at commit c721a23.

• This patch fails Spark unit tests.
• This patch merges cleanly.
• This patch adds no public classes.

[AmplabJenkins]

Merged build finished. Test FAILed.

[AmplabJenkins]

Test FAILed.
Refer to this link for build results (access rights to CI server needed):
https://amplab.cs.berkeley.edu/jenkins/job/SparkPullRequestBuilder/32546/
Test FAILed.

[nchammas]

@mengxr I'm gonna start PySpark performance tests for before/after this PR now.

[mengxr]

I'm closing this PR after discussing with @davies. useMemo is useful to avoid repeatedly serializing class objects. However, other than this case, it might bring overhead for Spark use cases. @davies is going to test a hybrid version that only memorizes class instances.