bump our fork version

[alyssafrazee]

Keeping the version of this fork the same as the original tidypredict fork (and possibly as a suitable version for the CRAN tidypredict) made it so our new confetti installation didn't realize it should replace our previously-installed version with this one. This PR changes our version (by a tiny tiny amount!! lol) from the dev version / main fork of tidypredict, and I'm going to see if I can peg a version in the confetti DESCRIPTION to ensure this one gets installed.

This PR also makes one last change to our fork, which is, when creating a decision tree SQL query, to round the outcome of the query to 3 decimal points. This makes it so instead of doing scoring by using a query like

select p_turnout as case when age > 50.54234234324234 then 0.7923842384723847 else 0.2123129387891273 end

it does

select p_turnout as case when age > 50.54 then 0.792 else 0.213

(the rounding of the age cutoff in this example is my original impetus for this fork and the difference in the tidypredict:::get_xgb_case function I look for; the rounding of the outcome probabilities to 3 decimal places is the 3 in this PR. This all helps us keep our queries under the bigquery character limit!)