BP4 streaming tests have timing heisenbug

[NAThompson]

Message:

406: [----------] Global test environment tear-down
406: [==========] 1 test from 1 test case ran. (171 ms total)
406: [  PASSED  ] 1 test.
406: unknown file: Failure
406: C++ exception with description "ERROR: File Staging.ZFPCompression.3x5.BPS.BB.BP4_stream was found with an index file but md.0 has not contained enough data within the specified timeout of 10.000000 seconds. index size = 704 metadata size = 0 expected size = 13535. One reason could be if the reader finds old data while the writer is creating the new files.: unspecified iostream_category error" thrown in the test body.
406: [  FAILED  ] CommonReadTest.ADIOS2CommonRead1D8 (10009 ms)
406: [----------] 1 test from CommonReadTest (10009 ms total)
406:
406: [----------] Global test environment tear-down
406: [==========] 1 test from 1 test case ran. (10010 ms total)
406: [  PASSED  ] 0 tests.
406: [  FAILED  ] 1 test, listed below:
406: [  FAILED  ] CommonReadTest.ADIOS2CommonRead1D8
406:
406:  1 FAILED TEST
406: unknown file: Failure
406: C++ exception with description "ERROR: File Staging.ZFPCompression.3x5.BPS.BB.BP4_stream was found with an index file but md.0 has not contained enough data within the specified timeout of 10.000000 seconds. index size = 704 metadata size = 0 expected size = 13535. One reason could be if the reader finds old data while the writer is creating the new files.: unspecified iostream_category error" thrown in the test body.
406: unknown file: Failure
406: C++ exception with description "ERROR: File Staging.ZFPCompression.3x5.BPS.BB.BP4_stream was found with an index file but md.0 has not contained enough data within the specified timeout of 10.000000 seconds. index size = 704 metadata size = 0 expected size = 13535. One reason could be if the reader finds old data while the writer is creating the new files.: unspecified iostream_category error" thrown in the test body.
406: [  FAILED  ] CommonReadTest.ADIOS2CommonRead1D8 (9997 ms)
406: [----------] 1 test from CommonReadTest (9997 ms total)
406:
406: [----------] Global test environment tear-down
406: [  FAILED  ] CommonReadTest.ADIOS2CommonRead1D8 (10000 ms)
406: [----------] 1 test from CommonReadTest (10000 ms total)
406:
406: [----------] Global test environment tear-down
406: [==========] 1 test from 1 test case ran. (9998 ms total)
406: [  PASSED  ] 0 tests.
406: [  FAILED  ] 1 test, listed below:
406: [  FAILED  ] CommonReadTest.ADIOS2CommonRead1D8
406:
406:  1 FAILED TEST
406: [==========] 1 test from 1 test case ran. (10000 ms total)
406: [  PASSED  ] 0 tests.
406: [  FAILED  ] 1 test, listed below:
406: [  FAILED  ] CommonReadTest.ADIOS2CommonRead1D8
406:
406:  1 FAILED TEST
406: unknown file: Failure
406: C++ exception with description "ERROR: File Staging.ZFPCompression.3x5.BPS.BB.BP4_stream was found with an index file but md.0 has not contained enough data within the specified timeout of 10.000000 seconds. index size = 704 metadata size = 0 expected size = 13535. One reason could be if the reader finds old data while the writer is creating the new files.: unspecified iostream_category error" thrown in the test body.
406: unknown file: Failure
406: C++ exception with description "ERROR: File Staging.ZFPCompression.3x5.BPS.BB.BP4_stream was found with an index file but md.0 has not contained enough data within the specified timeout of 10.000000 seconds. index size = 704 metadata size = 0 expected size = 13535. One reason could be if the reader finds old data while the writer is creating the new files.: unspecified iostream_category error" thrown in the test body.
406: [  FAILED  ] CommonReadTest.ADIOS2CommonRead1D8 (9993 ms)

Low priority, obviously, but fixing this issue would make all of the non-SST tests ASan and UBSan clean.

[chuckatkins]

These are unrelated to the sanitizers and are a known heisenbug in the BP4 staging tests. They crop up pretty regularly but the current ctest "retry up to 5 times until pass" option used in CI tends to hide it. We enable it in public CI because many of the staging and MPI tests timeout through no fault of their own really but instead from the massive variability in timing seen in cloud hosted CI because they're running in such a resource constrained VM and on heavily oversubscribed servers.

You can see these crop up in the regular runs @eisenhauer is doing to root out the heisenbugs where runs with the inverse setting of "retry up to 5 times until fail": https://open.cdash.org/index.php?compare1=61&filtercount=1&field1=revision&project=ADIOS&showfilters=0&limit=100&value1=2d678d0c057a0973b986054bf1f6fa5dfac30c7b&showfeed=0