Fix WatchQueueReader cancelling jobs too early

[J-N-K]

Related to #2727

While investigating the issue above I found that the new version of the WatchQueueReader erroneously cancelled the job when removing it from the stored jobs.

This is done on two occasions:

• If a jobs is stored and we try to replace it. In this case cancellation is correct.
• If the job executes. Cancelling the job here is wrong. It is not noticed by the tests because it interrupts the job, which is not checked for our short-running jobs. In the issue above the job is checking if the thread is interrupted and therefore fails.

@wborn I also moved the tests to the org.openhab.core package and added the JavaTest as you suggested.

Signed-off-by: Jan N. Klug github@klug.nrw

[wborn]

It looks OK to me! 👍 But I'd also then use .equals instead of == on line 255:

openhab-core/bundles/org.openhab.core/src/main/java/org/openhab/core/service/WatchQueueReader.java

Line 255 in e584507

if (kind == OVERFLOW) {

[J-N-K]

Done.