Usr iface II

[suhlatwork]

This pull request is supposed to replace #86. It is based on the changes of this pull request including commit 26e34ae1428d00dc559569474e8f2fa72143fb4c.

Compared to this the following changes have been done:

• add an accessor to the TTree pointer of an amplitudeMetadata object (similar to eventMetadata)
• use (pointers to) amplitudeMetadata objects in a couple of places instead of pointers to TTree and wave names:
  • ampIntegralMatrix::integrate
  • pwaLikelihood::addAmplitude
• move the several legacyScripts directories to the top level directory legacy, the idea would be to have stuff that still compiles directly in this directory, and stuff that does no longer compile in the same structure as in the current main directory, but below legacy/
  • add back C++ executables calcAmplitudes, calcIntegrals, and the two pwafits using the new file formats and facilities (except the file manager (obviously)), this should ease the migration a bit in case where some parts in the file manager are still missing
• update getReflectivity to work with old and new key file naming convention, as this is still required in the fitResult
• pwaLikelihood::init now requires a vector of tuples of wave name, wave description and threshold (instead of two separate vectors), the wave name has been added, it is required if the name of the on-file object does not correspond to the name constructed from the decay topology (this seems to be the simplest way to keep the possibility of having the same waves with different isobar parametrization (though I did not try this till the end))
• once again _decayAmps was resized for each new amplitude
• entirely (I hope so) remove some stuff no longer consistent with the new user interface:
  • binary amplitude files
  • old key file name convention
• similarly to calcAmplitudes allow to limit the bins for calcIntegrals
• check that a valid event type was specified for calcIntegrals
• do not stop if a integral file already exists, but simply skip (as done for the amplitudes)

Quite a number of smaller bugfixes and cosmetics:

• added include guards
• increase similarity between ROOT and NLopt fitter
  • add option to not calculate and store Hesse matrix for ROOT
  • add Cauchy priors for ROOT
  • check analytic Hesse matrix without storing it for NLopt
• remove further usages strings for amplitude and integral (names for ROOT objects) and replace by the static variables of the according meta data class
• some stuff (mostly cosmetics) lost in the rebases
• build system:
  • treat some more libraries/subdirectories in the new standard way
  • sort the various include paths
• increase ClassDef version of eventMetadata

I basically reshuffled all the commits, most commits should compile (again), 1504fca4be3e53d94f8060847fc1cd7c73029848 is the last commit largely equivalent to the current master.

I tested this branch on a single bin of my data obtaining the same result as with the current master, I do not know how to test this more properly (in any case some of the test scripts were broken?). This is a comparison of the memory usage:

	master (binary amplitudes)	master (ROOT amplitudes)	branch C++	branch py
calcAmplitudes	165 MB	415 MB	415 MB	420 MB
calcIntegrals	6 MB	1,45 GB	1,75 GB	1,75 GB
pwafit (ROOT)	120 MB*	530 MB	376 MB	396 MB

* during writing up to 350 MB

So calcIntegrals still needs to be improved, but that was already the case with ROOT amplitudes previously. A couple of issues raised in #86 are still open, but they can also be attacked later.

[suhlatwork]

Small update:

• move default arguments from the implementation to the declaration for pwaFit high-level interface
• do not have default arguments for the implementation of the function called by Boost for calcAmplitude, the default arguments are defined when adding this function to the Boost Python interface.
• do not have wrapper functions for the pwaFit classes, with the final interface Boost can figure this out automatically
• improve the extraction of a Boost tuple from a Boost Python tuple
• make both CovarianceMatrix methods of pwaLikelihood callable, previously always the one taking a TMatrixT was chosen.

The new commit largely equivalent to the current master is 1d1e964084dc9368830084ba7d142da0e102809a.