Test does not pass yet.

* remove dependency on config
* make checks optional
* support HS256

and make everything truly optional.

* Improve pubkey not found error handling

When the public key identified by the {Alg, KID} tuple is not found on
the IAM keystore server, it's possible to see errors like:

([email protected])140> epep:jwt_decode(SampleJWT).
** exception error: no function clause matching
                    public_key:do_verify(<<"eyJ0eXAiOiJKV1QiLCJhbGciOiJSUzI1NiIsImtpZCI6IjIwMTcwNTIwLTAwOjAwOjAwIn0.eyJpc3MiOiJodHRwOi8vbG9jYWxob3N0OjEyMzIx"...>>,
                                         sha256,
                                         <<229,188,162,247,201,233,118,32,115,206,156,
                                           169,17,221,78,157,161,147,46,179,42,219,66,
                                           15,139,91,...>>,
                                         {error,not_found}) (public_key.erl, line 782)
     in function  jwtf:public_key_verify/4 (src/jwtf.erl, line 212)
     in call from jwtf:decode/3 (src/jwtf.erl, line 30)

Modify key/1 and public_key_not_found_test/0 to account for keystore
changing from returning an error tuple to throwing one.

Tolerate 5 crashes per 10 seconds

Improve restart strategy

Implement jwtf:encode/3 for encoding JSON Web Tokens. Test encode/decode
round trip for each supported alg.

Implement encode

Simplify tests

Move key cache to epep

Currently jwtf tests don't run in a continuous integration
environment, presumably due to dependency rules.

This splits the tests into their own module, but requires exposing a
couple new functions in jwtf to support them. Some long lines were
also broken into smaller lengths.

Separate tests into dedicated module

Port reduce_false.js and reduce_builtin.js to Elixir

It can be desirable in some cases for decoded JSON to e.g. return
maps instead of the default data structure, which is not currently
possible.

This exposes a new function `couch_util:decode/2`, the second
parameter being a list of options passed to `jiffy:decode/2`.

I found a .view file with a db_header in production (cause unknown but
I'm hoping it's manual intervention).

This patch means we'll reset the index if we find something other than
a view header when looking for one.

This reverts commit 91ecf67.

Port changes.js test suite into elixir

…2.0-disable-legacy-checks

Upgrade credo to 1.2.0 disable legacy checks

There's a race between the meck:wait call in setup and killing the
config_event process. Its possible that we could kill and restart the
config_event process after meck:wait returns, but before
gen_event:add_sup_handler is called. More likely, we could end up
killing the config_event gen_event process before its fully handled the
add_sup_handler message and linked the notifier pid.

This avoids the race by waiting for config_event to return that it has
processed the add_sup_handler message instead of relying on meck:wait
for the subscription call.

…st_to_elixir

Port form_submit.js test to Elixir

doc: link README-DEV in README

…lixir (apache#2596)

* Port _design_docs tests from js to elixir
* Port design_options and design_paths tests from js to elixir

After mango python tests are run, a bunch of dbs are typically left
around, e.g. `mango_test_048b290b574d4039981893097ab71912`

This deletes those test dbs after they are no longer in use.

(cherry picked from commit e05e3cd)

Tests based on class `UsersDbTests` don't clean up the user docs it
puts in the `_users` db. This uses the classmethod `tearDownClass` to
delete those docs.

(cherry picked from commit 3d559eb)

This primarily wraps long lines and removes trailing whitespace in the
README.md and TODO.md files. In `test/README.md`, it updates the default
admin username and password used by `dev/run`.

…-editable-ini-fix

Move "users_db_security_editable" to the correct location in the ini file

Recent changes in how `jiffy:decode/1` handles malformed JSON has caused
`jwtf:decode/3` to fail to properly return a bad request 400 response
for some malformed tokens.

First, this changes the name of the function to `decode_b64url_json/1`,
indicating that it decodes something that has been first been JSON
encoded, and then base64url encoded.

More substantially, it wraps both the base64url and jiffy decoding in a
try/catch block, since both can throw errors, while the former can also
return an error tuple. Tests have been added to ensure all code paths
are covered.

Upgrade credo to 1.3.0

Handle malformed tokens with jiffy 1.x

If a filtered changes feed hit a rewind we would send a bare `integer()`
value for the Seq. If this was used again during a rewind it causes a
competely rewind to zero due to not having the `node()` and UUID
`binary()` values to calculate a new start seq.

…winds

Send correct seq values for filtered changes

Closes apache#2655

Set cookie domain when DELETE'ing

Import jwtf

The "alg" check can now take list of algorithms that are supported,
which must be from the valid list of algorithms.

Enhance alg check

* Add support for specify a custom config file for CouchDB startup during testing
* Port proxyauth test from js to elixir

Throw if an unknown check is passed to jwtf:decode

Add JWT Authentication Handler

Co-authored-by: Robert Newson <[email protected]>
Co-authored-by: Joan Touzet <[email protected]>