The new option will run the semantic checks and then dump the parse tree
and all the symbols. This is equivalent to running the driver twice,
once with -fdebug-dump-parse-tree and then with
the -fdebug-dump-symbols action flag.
Currently we wouldn't be able to achieve the same by simply running:
flang-new -fc1 -fdebug-dump-parse-tree -fdebug-dump-symbols <input-file>
That's because the new driver will only run one frontend action per
invocation (both of the flags used here are action flags). Diverging
from this design would lead to costly compromises and it's best avoided.
We may want to consider re-designing our debugging actions (and action
options) in the future so that there's more code re-use. For now, I'm
focusing on making sure that we support all the major cases requested by