Multiple issues with `ConditionEvaluationResult` reason values

[sbrannen]

Overview

As I mentioned in #4699, there are few long-standing issues with ConditionEvaluationResult.

1. When a custom reason is supplied but a null value is supplied for the default reason to ConditionEvaluationResult.disabled(String, String), the generated reason is "null ==> custom"; whereas, ideally that should just be "custom".
2. We have always allowed null for any of the reasons used to create a ConditionEvaluationResult, but the Javadoc does not reflect that.
3. We should not use a blank reason as-is (such as an empty string or a string containing only whitespace). Instead, we should treat a blank reason the same as we do for a null reason.

Point 1 is a bug. Point 2 is a documentation issue. Point 3 is rather an enhancement. However, since these are all closely related, I have created this single GitHub issue to address all three together.

Related Issues

• ConditionEvaluationResult APIs should accept @Nullable reasons #4698
• Support @Nullable reasons in ConditionEvaluationResult APIs #4699

[sbrannen]

Reopening to backport to 5.13.x.

[sbrannen]

Reopening to backport to 5.13.x.

Backported in e5d0e0b.