Fix procdump

[lumpfish]

No description provided.

[lumpfish]

sanity.system test runs are in
jdk8
openj9
x86-64_windows: https://ci.adoptopenjdk.net/view/Test_grinder/job/Grinder/4541/ (two test failures but unrelated to this change)
jdk11
hotspot
x86-64_windows: https://ci.adoptopenjdk.net/view/Test_grinder/job/Grinder/4542/

[lumpfish]

Fixes #86

[Mesbah-Alam]

Could we please squash the commits into a single commit?

[Mesbah-Alam]

The PR looks good.

The Grinders above didn't really run the code added in this PR. It'd be conclusive to see a test where we have a "hang" and the procdump tool gets invoked successfully.

Running a simple grinder with the original failing test didn't give us the original "hang", as it's an intermittent issue.

So, I added a target for SampleOverrunProcess and ran it in grinder, and I do see the dump being created in the resultant zip.

Output from 1.CL.jvmdump.stdout:

[11:01:08] Dump 1 initiated: E:\workspace\Grinder\openjdk-tests\TKG\test_output_1604595588538\SampleOverrunProcess_0\20201105-105950-SampleOverrunProcess\results\java.exe_201105_110108.dmp
[11:01:08] Dump 1 writing: Estimated dump file size is 114 MB.
[11:01:08] Dump 1 complete: 114 MB written in 0.2 seconds
[11:01:08] Dump count reached.

I am not sure what exact Perl version (32 bit or 64 bit) was used in the build. It says:

"This is perl 5, version 30, subversion 3 (v5.30.3) built for x86_64-cygwin-threads-multi".

[lumpfish]

I would imagine 32 bit perl would say "built for x86_32-cygwin-threads-multi", and it would be included in a 32 bit cygwin installation.
In my local testing I used Strawberry perl (native Windows version of perl).
Presumably the use of 32 bit will diminish over time. Strawberry perl are still building a 32 bit version: http://strawberryperl.com/releases.html.

[lumpfish]

No reason not to squash. Using github's "Squash and merge" is easiest.

[Mesbah-Alam]

LGTM