Fixes `render_to_object()` for multi-output recipes

[schuylermartin45]

• Function now handles rendering multi-output recipes. This is accomplished by handling Collection Nodes as a special case. - This code isn't perfectly recursive, but it gets the current job done. - This fixes 105 previously unparsed recipes
• Modifies an existing regex to dramatically boost recipe compatibility
• Adds unit tests
• Adds script/ directory - This directory contains tools for developers that should not be broadly published - Adds analyze_aggregate_errors.py script to help identify the biggest sources of recipe parsing errors.

Here are the current error counts from aggregate, courtesy of the new script:

==== Histogram ====
{
  "'cli_opts'": 1,
  "'dict' object has no attribute 'append'": 8,
  "'int' object does not support item assignment": 1,
  "'outputs'": 1,
  "'requirements'": 1,
  "'str' object does not support item assignment": 11,
  "'str' object has no attribute 'append'": 2,
  "(PosixPath('/Users/smartin/work/aggregate/clangdev-feedstock/recipe'), 'can only concatenate str (not \"int\") to str')": 1,
  "(PosixPath('/Users/smartin/work/aggregate/gtk3-feedstock/recipe'), \"unexpected char '#' at 4499 (at line 133)\")": 1,
  "(PosixPath('/Users/smartin/work/aggregate/theano-feedstock/recipe'), \"render.<locals>.<lambda>() got multiple values for argument 'max_pin'\")": 1,
  "expected '<document start>', but found '<scalar>'": 2,
  "found undefined alias 'test'": 1,
  "list index out of range": 54,
  "list indices must be integers or slices, not object": 1,
  "list indices must be integers or slices, not str": 4,
  "mapping values are not allowed here": 1,
  "pop from empty list": 11,
  "string indices must be integers, not 'str'": 1,
  "while constructing a mapping": 11,
  "while parsing a block collection": 1,
  "while parsing a block mapping": 3,
  "while parsing a flow mapping": 12,
  "while scanning a block scalar": 1,
  "while scanning a quoted scalar": 4,
  "while scanning an alias": 4,
  "while scanning an anchor": 1,
  "while scanning for the next token": 54
}
==== Stats ====
Total error count: 194

[schuylermartin45]

As of the last commit, the error count is now down to 166, after a quick win.

[schuylermartin45]

I think the regex change is the only thing being left out.

Realistically for that I'd like to eventually run the parser against every recipe in aggregate and try to keep track of parsing regressions/maintain some minimum acceptance level. I guess that could be an integration test of sorts, let me capture that as a future ticket.