As the number of binding schemas has grown, the time to run
dt_binding_check has gotten pretty slow. A large part of this is due to
the slow startup time of Python (a well documented problem). There's not
currently any benefit to running dt-doc-validate one file at a time, so
let's switch it to run a single rule. Doing this means we loose the make
parallelism, but we can use xargs instead. This speeds up the validation
time from several minutes to <10 sec.

Since the validation is a single step with no output, we move running it
as part of the processed-schema-examples.json target. We also need to
reorder the extra-y entries so the validation is run first rather than
after all the examples are extracted.
Signed-off-by: Rob Herring <robh@kernel.org>