Add thread type feature

There are two thread types, discussion and question. Discussions have
the same behavior that current threads do. When a question is returned
from the API, it contains separate fields for its endorsed and non-
endorsed entries.

Co-authored-by: Jim Abramson <jsa@edx.org>

Editing a comment such that it matched a blocked hash would cause the
API to return 503 but actually update the comment. This is now fixed.

Refactor search, fix a bug, and implement spelling correction

If a search has no results, then we will get suggestions from
Elasticsearch, re-query with the top suggestion, and return the results
from that query along with the suggestion that was used.

This change also changes search so that a document must match all terms
to be included in the results instead of any term.

This contains several improvements and bug-fixes:

* Take advantage of indexing changes in ead2f5e3 to reduce the number of
  round trips to execute a search
* Use Tire scan to include all matching threads
* Execute final sort/pagination in Mongo
* Share sorting code
* Eliminate separate presenter for search results (which removes unused
  highlighting functionality)

JIRA: FOR-607

use a single index in elasticsearch for both document types.

WARNING: Must rebuild search index!

Starting with this change, the name of the search index used by the
service has changed.  The new search index must be created before
deploying this version of the application.  There are two ways of doing
this:

* Offline (recommended)
Follow [these instructions](https://github.com/edx/cs_comments_service/wiki/Search-Indexes#offline-procedure) to perform an offline rebuild.

* Online
1. Deploy this code to a host which is not serving HTTP requests, and
run `rake search:rebuild`
2. When the rebuild finishes, deploy the updated code on live servers.
3. run `rake search:catchup[MINUTES]` where minutes is the amount of
time it took to run rebuild in step 1.
4. run `rake search:prune`

Fix task seed in Rakefile

Revise/add rake tasks for search index mgmt.  Bump Tire to 0.6.2.

Tire is being upgraded in order to get the latest import-related fixes, such as:

https://github.com/karmi/retire/commit/13abc5b351a66301e246adf9e657fd3b5e9c66aa
https://github.com/karmi/retire/commit/45eb4f988ddc51bd51236e0c918804dc3ee29a9b
https://github.com/karmi/retire/commit/7117550ac76f849dd8c55cb1f2bf24a68de40d7c
https://github.com/karmi/retire/commit/67b89035cf7e63b87f6656eda6e73be500f4977a

Return total result count for text search queries

add create/update dates to comments index, for use in statistical facets

add migrate-index.sh for simplified reindexing.

E0d/stream2es

Index and query ES using english analyzer

fixed typo

removed refactor leave behinds

missed push

cr comments

removing extra scheme part

fixed typo, clarified comment

Remove email from User model and API

It was never actually used for anything except to enforce uniqueness,
which should be done by the LMS.

upgrade tire to 0.5.2, update use of deprecated "text" search api

its value is always true regardless of cluster health, and it is removed
in ES v1.x.

Fix active threads endpoint

Previously, pagination was done before reducing from content to threads,
which meant that fewer threads than requested could be returned and that
the number of pages returned could be incorrect.

Return 404 instead of 400 when thread not found

Co-authored-by: Greg Price <gprice@edx.org>

don't require api key for heartbeat.

* don't require api key for heartbeat, nor expose tracebacks.
* add tests for heartbeat.
* fix test with hard tabs.

JIRA: FOR-96

reinstate disconnection from mongo in unicorn master thread

https://github.com/edx/cs_comments_service/commit/e61430d9438abf1db64fcac0eb9dd3e0f18a07bd was inadvertently lost in a recent change, this puts it back.

add heartbeat endpoint