Make KeepDropSwitch fail earlier on invalid inputs

[tmadlener]

BEGINRELEASENOTES

• Check the commands for the KeepDropSwitch in the constructor instead of re-parsing them every time isOn is called.
• Move the KeepDropSwitch into the k4FWCore namespace and deprecate the unnamespaced usage.
• Make the IOSvc fail in initialize in case the outputCommands are invalid
• Add unittests for KeepDropSwitch

ENDRELEASENOTES

Not sure if the deprecation of the global namespace definition of KeepDropSwitch is necessary or if we can simply remove the deprecated definition immediately.

Fixes #348

[tmadlener]

A quick benchmark shows that the interesting use case of constructing the KeepDropSwitch and running isOn on a number of collections once that this is now twice as fast for the default case of "keep *", and that runtime is roughly stable with multiple output commands, whereas the existing implementation exhibits some visible scaling behavior in the number of commands.

It is not the most performance sensitive piece in the whole chain, but I think it is still nice to see that the pre-processing we do in the constructor pays off for the overall runtime.

[jmcarcell]

Yes, I don't think any difference will be noticeable, at least I never saw it in profiles. I think the unnamespaced one can be removed, I don't think it is used anywhere else.

[jmcarcell]

I'm not sure for how long Gaudi will be buildable with GCC 11 since it has been a while they only test with GCC 13 and above.