* initial type filtering bar

* use filter for items in collection landing

* filter search by type

* remove redundant heading

* prettier

* 1/3 2/3 red third blue third

* handle no non collection items state

* prettier

* merge fix

* don't drop existing query items when changing type

* move personal items on root

* use modal and proper route for collection edit

* breadcrumb hover state

* redirect to parent collection on archive

* stay on current collection when creating child collection

* return to previous url when canceling pulse creation

* hide reset button on forms by default

* lint fix

* add collection filter for search

* return and show display in card items for collections

* dont log

* use gear instead of person

* test fixes

* cr

* fix typo / expectation

There was code to remove the special_type from the field before
returning the metadata. That code was in place so that the frontend
doesn't treat that field like a real PK/FK (since it's a nested
query). That special_type was being removed before the code to added
dimension options was ran, causing it to add dimension options for
PK/FK values (which it shouldn't). This commit just separates out the
code that clears the special type so it can run after dimension
options are added.

To support this Cards now include fingerprints in the
result_metadata. That fingerprint is included in the query dictionary
and is available to the query pipeline. The expand/resolve/binning
middleware can make use of that fingerprint data to build the binned
query. Dimension options for these kinds of queries will also now be
available.

Note that binning is only available on questions that have result
metadata. If no result metadata (and thus no fingerprint) is
available, no dimensions options will be included and an exception
will be thrown if a binned query is attempted. In most cases, just
running a card once will populate that data.

Originally this work was done via f49e32a6 and 7977fd45, but those
commits were reverted as some test fixes were needed on a few drivers.

Previously the driver was only looking one and two levels deep for the
source table. In the case of nested and binned queries, this could be
buried deeper in the MBQL query. This commit switches form an `or`
clause and a couple of `get-in`s to a `postwalk` that will find the
table however it's nested in the query.

* fix all user collection crumb spacing

* new browse link

* change label

* show regardless of pinned items

* homepage blocks

* table x-ray and reference links

* overworld db links

* minor copy changes

* move personal collections

* copy changes

* remove GSG from userland

* change default collection name

* more updates

* microcopy cleanup

* prettier

* fix colors

* fix typo

* rm overworld links

* lint

This commit stores fingerprints and special type information for the
results of cards that have native queries, similar to what we store
for MBQL queries.

Google and LDAP users shouldn't be able to change their own email
addresses as we get that from the source.

To support this Cards now include fingerprints in the
result_metadata. That fingerprint is included in the query dictionary
and is available to the query pipeline. The expand/resolve/binning
middleware can make use of that fingerprint data to build the binned
query. Dimension options for these kinds of queries will also now be
available.

Note that binning is only available on questions that have result
metadata. If no result metadata (and thus no fingerprint) is
available, no dimensions options will be included and an exception
will be thrown if a binned query is attempted. In most cases, just
running a card once will populate that data.

Fixes #5938

The pulse GET responses include a PulseCard that isn't exactly a
PulseCard. It includes some extra information from the underlying card
useful for the FE to include in the display. When adding a card to an
existing pulse that has one or more cards already, the PUT include a
mixture of these HybridPulseCards (the existing cards for the pulse)
and CardRefs (for newly added cards). This commit changes the schema
to allow either and updates the tests accordingly.

This also adds similar flexibility to the POST so that it is
consistent with the PUT.

Fixes #7789

Recently the `/collection/:id` endpoint was updated to only return
collection metadata and separately (via the `/collection/:id/items`
endpoint) return the children of the collection. This updates the test
to follow that same split.

The test query to ensure alerts aren't returned with pulse search, and
just filtering out the id of the pulse to ensure it's not in the
results. That test can fail if there happens to be a search result
with the same id as the (missing) pulse. Adding another predicate to
ensure it's only looking for pulses.

Anytime a new card/dashboard/pulse is created with a collection
position, or a card/dashboard/pulses has it's collection or collection
position updated, it's possible that other items in the collection
will need to have their collection position adjusted. This commit adds
a check to the PUT and POST routes of cards/dashboards/pulses to check
to see if the change impacts the position of other
cards/dashboards/pulses in the collection.

This also uses the same logic for the bulk card collection updates and
will adjust impacted cards as needed.

Fixes various collections UI bugs and eliminates duplicate API calls

These were coming back with pulses, needed to include another query
clause to remove them.