Avoid use of hard-coded password in Open Liberty configuration

[scottkurz]

See: #335 though the fix there broke the project tests, so was reverted.

[Emily-Jiang]

Back to the drawing board. The previous PR #336 did cause JWT tests to fail though. A different solution needs to be provided.

[scottkurz]

It seems like there could be two directions to go in, to enable the Open Liberty config change made in #336.

1. Have the automated test generate the certificates using the password generated by the Open Liberty runtime. It looks like they might be generated here https://github.com/eclipse/microprofile-starter/blob/master/gencerts.sh but this might be done with a static, fixed key using "atbash" as password. This might require some rework/reordering of the test logic. I'm not too familiar with any of this. ((I noticed too it was: https://github.com/eclipse/microprofile-starter/blob/master/src/main/resources/files/TestSecureController.java.tpl where the JWT code loads the keys.)

2. Accept that the Open Liberty server config during the test is going to be different from that generated from the MicroProfile starter, and accept that that is "close enough" or acceptable. In this approach, we're OK taking the config actually generated and overriding it using a configDropin or similar override mechanism to test using something like the existing config, with the pre-generated certificate.

But in either case, I'm suggesting adjusting the tests and using the Open Liberty runtime config I suggested from #336.

Hope that moves the discussion forward a bit.

[gkwan-ibm]

and use p12 instead of JKS

[Karm]

@Emily-Jiang Is this still an issue?