convert functions to binned format as additional output

[hftsoi]

for each candidate function, add an extra column to candidate_functions df that presents the smooth function back in binned histogram, like the input y. this can be useful for use cases where a smoothed histogram is desired over an explicit function. also add plotting for this.

[hftsoi]

oh for hep analyses, it can be useful for converting a bkg histogram with limited statistics to a smoothed version-- like qcd in stat limited signal region, which has relied on deriving an additional template with looser cuts and then scale it back to normal selection-- this can be done by symbolfit without needing a separate template..

[hftsoi]

the above can be done for all cases as default. but can add an additional option that, instead of fitting to the given binned data, first converts the input binned data into a density, fit it to get a smooth density, then integrate to get back the binned data but as a smoothed version. this fitting-to-density should be enabled explicitly through the option and the default fit should be disabled simultaneously, otherwise two different datasets will be fitted upon a single fit() call.

[hftsoi]

this is mostly for bkg estimation scenario in binned analysis, but this method would uncorrelate uncertainties in the smoothed histogram from the analysis...