Merge IFunctionalTest::analyzeResults() into IFunctionalTest::runTest() [Tests are a sequence of actions, each of which may result in failure. Therefore there is no natural division between test and analysis]. Also detect errors from Python script execution commands. This partly resolves #1571.

The remaining two (in SubtestRegistry and OutputDataReadFactory) are
probably best handled by a somewhat deeper refactoring.

Further decoupling in the futest machinery: Sample builders need not to know about FutestSuite (aka service), but only about the subtest_item (of type IParameterized*).