Bayesian average method for one fit function added to juobs_tools.jl

update in getall_eigvals and uwerr(ydata) removed in juobs_tools.jl/fit_routine in order to choose a different wpm windows. Otherwise it does not work

dec_const uses time reflected correlators

automatic truncation t0 + unfinished runs

format modifications in doc

get_chi2 + fit_routine restructured 

fit_routine uses Optim
LsqFit deps removed.

functions that calls uwerr now includes wpm

cdata = [[data_r1[k], data_r2[k]] for k=1:length(data_r1)]
is no longer needed