Skip known errors for long-running format stress sanitizer tasks

The new Evergreen task format-stress-sanitizer-test failed consistently ever since being migrated from Jenkins (WT-5000). The new model in Evergreen is to run the test for ~24 hrs to simulate the continuously-triggered testing pattern we used to have in Jenkins. 

In one of the failed task runs, the test ran through 24 hrs (as expected) and hit a few instances of "heap-use-after-free" (WT-5219) and cache stuck failures (WT-3694, WT-5370 and friends).

https://evergreen.mongodb.com/task/wiredtiger_ubuntu1804_format_stress_sanitizer_test_5624464bba22fd9876fecf412c8eaa2e009eab85_19_12_27_02_39_23

We need to either address those known failures (which could be hard esp. the cache stuck failures), or find another way to avoid consistently failing the task. 

cc: Haseeb Bokhari Alexander Gorrod