[FIXED JENKINS-15587] Do not issue a warning about unparsable build timestamps if it is just a number, i.e. failed symlink.

Not really solving anything here—there is no plausible implementation that could run under these circumstances—
but at least reporting the problem more clearly.

[JENKINS-16276] and various fixes for related exceptions

Otherwise it might try to access a Descriptor that's not yet registered.

Happens when build has no parent:

15:06:14 FATAL: null
15:06:14 java.lang.NullPointerException
15:06:14    at hudson.matrix.MatrixBuild$MatrixBuildExecution.doRun(MatrixBuild.java:371)
15:06:14    at hudson.model.AbstractBuild$AbstractBuildExecution.run(AbstractBuild.java:580)
15:06:14    at hudson.model.Run.execute(Run.java:1575)
15:06:14    at hudson.matrix.MatrixBuild.run(MatrixBuild.java:304)
15:06:14    at hudson.model.ResourceController.execute(ResourceController.java:88)
15:06:14    at hudson.model.Executor.run(Executor.java:237)
15:06:14    at hudson.model.OneOffExecutor.run(OneOffExecutor.java:66)

Suppressing some known-good instances of an idiom otherwise warned about by NetBeans plugin for Jenkins.

Fix polish translations

Bump mailer version to 1.5

[FIXED JENKINS-15652] “Solving” errors by just return null from ALLRM.search and reducing to one-line warnings.
@kohsuke confirms that the intent of throwing AIOOBE was to track down definite code errors;
but in at least some known occurrences of this bug, the problem is invalid data:
build records on disk with <number>s that are duplicated and/or out of order.
Since that violates the precondition of the class, we do not pretend to successfully search for such records.
At some point it would make sense to have the code detect anomalous conditions and initiate a builds folder cleanup.
Take only the last build of a given number, and discard any earlier with higher number.
I.e. just set <number> to 0 or something and ignore such builds, or rename build.xml to build-broken.xml.