Before the ansible refactor, we were copying and pasting a lot of
ansible configs.  Instead of copying and pasting the large tar balls
that went with those deployments, they all lived in a central
'thirdparty' directory.  However, that creates confusing ansible
playbooks and directory structure, and it's much simpler to have things
co-located.  This resolves that change.

Nothing massive here, mainly just making the structure a bit deeper by
creating top level directories for the base install and the
competitions, then spliting up the docs a little.

Because roles are getting centralized, the documentation needs to be
updated with how and how to configure it.

This adds a centralized, refactored, etl pipeline that all competitions
can depend on.  This prevents the rampant copying and pasting of etl
pipelines as new competitions are created, and prevents the coderot from
modifying later etl pipelines that don't make it back into the original
ones.

Version 1 because this is based on the Climate2030 pipeline that doesn't
have a number of features that older, more complete pipelines have.
Those will get added incrementally.

Issue #53: Refactor ETL code to create central utilities

(Frank found the correct text in commit 4c3dd025.)

Also updates various docs for the system as it came to be rather
than the system as we originally intended it.

The DESIGN.md came from github.com/OpenTechStrategies/torque
before we split it apart

We're not Wikipedia, but we can try to be as tasteful as Wikipedia:

https://en.wikipedia.org/wiki/Wikipedia:Manual_of_Style/Linking#Duplicate_and_repeat_links

This reverts part of commit 92ce78fa, which overly-eagerly moved
the entire top-level README.md into the 100Change2020/etl/ subdir.
While it is true that most of that README.md was specific to that
competition's ETL process, the first section was a high-level
explanation of this overall repository, so that's what I've restored.

A lot of the installation and running instructions were based on the
2016 version of the project, as well as the pre-torque csv2wiki based
world.  This has been all brought up to date.

The updates in this include:

* Move fix-csv to compose-csvs
* Move csv munging code out of wiki-refresh base script into python, now
  located in compose-csvs
* Call csv2wiki via python3 module loading
* Update macfound-internal-csv2wiki-config.tmpl to include judge data
* Update macfound-internal-csv2wiki-config.tmpl to do embed youtube
  formatting

More generally, this update takes three different csvs and puts them
together into one set of data that's culled by which proposals have been
reviewed favorably, then adds judge review data to those proposals to
output to the wiki instance.  This work is all done in one place, with
wiki-refresh now being a small combiner script with variables about data
placement.

After all CSV processing, copy the deliverable CSV to a final name,
"processed-100andchangeExport-all-judges.csv", that will not change if
we add more processing stages in the future.  Also, update and improve
documentation in README.md and comments in wiki-refresh.

The script comments note that the user should feel warned, so we
should make the warning explicit.

Also, note the new beautifulsoup dependency in README

This reverts commit 67bd3eb2.

Underscore is discouraged only in package names, not in module names,
in https://www.python.org/dev/peps/pep-0008/#package-and-module-names.
In any case, our change had switched *to* underscore, not away; if we
were going to take the advice we had thought the PEP was giving, we
would have called the thing "fixcsv" :-).  But let's just revert
entirely.  This isn't a package anyway; it's at most a module, and
hyphen is more common in filenames (and usually easier to type).