fixes LBCORE-1229

See merge request !231

NamedMeasurements (user defined entries in CTest XML output) must already contain
escaped XML to work, so no need to escape them again.

This reverts commit 05538d04.

- fixed warnings from clang
- fixed issues with ROOT 6.08
- minor polishing
- minor fixes to tests
- fixed GAUDI-1221

See merge request !230

CTest output parser already escapes special characters for XML, and the escape in the
Gaudi test runner was resulting in double escaping.

Since zero can be a valid number for various EventID related parameters,
e.g. run# or event#, we need invalid defaults to be able to distinguish
between EventIDs that are set with lumiblock info, vs run# info

See merge request !199

- Increase test coverage and add performance check to test_GaudiTiming
- Remove unnecessary calls to getpid (saves ~10ns per query() call):
    - Remove getpid calls in case pid is -1 (self)
    - Remove ugly define that was hiding the getpid() calls in the
      source code
- Switch from getrusage() to times() (saves ~100ns per call)
- Only create ProcessHandle for Windows
- Add templated version of currentTime() so it can be inlined
- Allow System::getProcess to be inlined (move Windows code to
  constructor)
- Remove private caches for query() calls as those were not used

See merge request !228

See merge request !229

Add support for copy assigning or constructing
`XYZHandle<const T>` from a `XYZHandle<T>`. 

Extended the test `MyGaudiAlg.py` test for this feature
and `ToolHandle<const T>` support in general.

Developed due to the discussion in !215

See merge request !226

To fix an issue with Atlas in !215

See merge request !225

I would like to use ToolHandle more, but I also want to use it optimally efficiently, by which I mean I want inline access to the underlying tool pointer, with no validity checks (I always `retrieve()` my tools up front in initialize()) and true const behaviour (for thread safety). Essentially what I want is `GaudiHandle<T>::get()`, but unfortunately in the current implementation this method is hidden behind the virtual `ToolHandle<T>::get()` method, that also down casts the tool to `IAlgTool*` (so no use for direct usage).

This MR fixes both these issues. get() is now fully optimal in `ToolHandle<T>` as it simply refers back to GaudiHandle in an inline method.

The `BaseToolHandle<T>::get()` method is just a wrapper that calls a hidden (protected) virtual function (`getAsIAlgTool()`) that returns what get() previously did.

I also tightened up the const'ness, providing both const and non-const methods that return const and non-const tool pointers respectively.

I also updated a few methods to remove multiple return statements (which normally cause compilers to abort inlining) and adding some `noexcept` here and there..

The only API change is that `ToolHandle<T>::get()` now returns `T*` instead of `IAlgTool*`, but this is not a problem as the user can still choose to use it as a `IAlgTool*` if they want..

Gaudi master builds fine with this, not tested any other projects yet.

cheers Chris

See merge request !218

- Increase test coverage and add performance check to test_GaudiTiming
- Remove unnecessary calls to getpid (saves ~10ns per query() call):
    - Remove getpid calls in case pid is -1 (self)
    - Remove ugly define that was hiding the getpid() calls in the source code
- Switch from getrusage() to times() (saves ~100ns per call)
- Only create ProcessHandle for Windows
- Add templated version of currentTime() so it can be inlined
- Allow System::getProcess to be inlined (move Windows code to constructor)
- Remove private caches for query() calls as those were not used

- fixed issue in CMake modules test
- fixed BrunelScenarioGraphScheduler
- clean up CPUCruncher

See merge request !227

make use intermediate nodes are created when objects are added to TS