Handle re-used variables

[schuylermartin45]

There are some recipe files that re-use JINJA variables to append to themselves. Because the variables table is a dictionary, this means that only 1 JINJA set line will ever be interpreted.

The most simplistic way to handle this case would be do a first-pass-reduction and combine all these variables into 1 long JINJA line. Given how rare this pattern is, that might be the best trade-off to make.

See here: AnacondaRecipes/streamlit-feedstock#44 (comment)

[schuylermartin45]

Annoyingly, it seems that some recipes use selectors on these lines to optionally append to a JINJA variable:
https://github.com/AnacondaRecipes/scipy-feedstock/blob/fad9fc0984e6fd527f47db7c4141def967a40540/recipe/meta.yaml#L44

[schuylermartin45]

Requirement: to move comments as well.