Add new mean method that uses all interations

[AugustT]

Add in @GPowney work on mean estimations using all iterations

[drnickisaac]

We discussed this together, There are basically two variants: 1) a geomean(occupancy) and 2) exp(mean(logit(occ))), i.e. geomean(odds). The first option is what @CharlieOuthwaite is using for her PhD "big paper analysis", the latter is what Gary is using for the 2018 pollinator indicator.

Both share a couple of features

• zero kludge: estimated occupacny of zero is reset to some small value
• requires complete time series

@GPowney to document this in the vignette. @CharlieOuthwaite to test option 1 against he own code.

[GPowney]

I've added a function into BRCindicators, CompositeTrends. @CharlieOuthwaite can you test this against your code, do you get the same results?!

[CharlieOuthwaite]

Just ran the "geometric_raw_occ" version on ants, the indicator values are the same. The CIs are slightly different but I think that might be due to how things have been rescaled as mine start at 100 in year 1.

[GPowney]

These methods have been added and documented in TrendSummaries. See here.