Fix integration classes to use both a relative or n absolute accuracy

as requested in https://savannah.cern.ch/bugs/?70853

Improve TF1::Integral to add the precision parameter. 
Add a TF1::IntegralOneDim where one can specify both epsrel and epsabs and it returns also the error
Clean up the TF1::Integral class to fix some ambiguities 
Remove the possibility to pass vector of parameters in Tf1::Integral

remove obsolate file minuit2/src/FitterUtil.cxx


git-svn-id: http://root.cern.ch/svn/root/trunk@47641 27541ba8-7e3a-0410-8455-c3a389f83636