Remove the per-project files so we don't have to keep them in sync.
GitHub's UI links to everything except the contributing guide, so add a
section about that to the readme.
There was a bug that came as the result of an early optimization done
within ID tracking that caused loading parameters to fail in a very
specific and rare edge case. That edge case only occurred when the
parameter was being set within all 3 standard branches of an if block,
since the optimization would assume that the parameter was never being
referenced and was only ever being set. This would cause the variable to
be set to undefined.
The fix for this was to remove the optimization and still continue to
load in the parameter even if it is set in all 3 branches.
This fixes a bug that existed because namespaces within `{% set %}`
were treated as a special case. This special case had the side-effect
of bypassing the code which allows for tuples to be assigned to.
The solution was to make tuple handling (and by extension, primary token
handling) aware of namespaces so that namespace tokens can be handled
appropriately. This is handled in a backwards-compatible way which
ensures that we do not try to parse namespace tokens when we otherwise
would be expecting to parse out name tokens with attributes.
Namespace instance checks are moved earlier, and deduplicated, so that
all checks are done before the assignment. Otherwise, the check could be
emitted in the middle of the tuple.
Add the phrases 'multiline comment' and 'triple quotes' to docs
in the templates/#block-assignments section. This allows for new
users to find this alternative easily.
