* Writing migration notes

* search events 1-1-0

* backwards compatible adjustment

* making context optional

* refactor: de-duplicate variable usage

* fix incorrect goal-line plotting in pow and log y-axis scales

* add Loki variants to validate and enforce desired behavior

* update loki snapshots

* Change run to make-run in process-query-for-card

Fixes #44160

There were two levers where a card could change how it was run.

`qp` this is basically `qp/process-query` with some optional userland
flags and constraints marked on the passed in query.
In the case of pivot tables in the
`/pivot/:card-id/query` endpoints it was `qp.pivot/process-query`

'run' this defaults to a streaming response that takes the `qp` param
and `export-format` and passes a query to qp. Most often this is
overriden to not stream the response.

Unfortunately, if `run` is overwritten, then `qp` is thrown away, and
the caller who overrides `run` must determine the correct `qp` to use.

However, when running `process-query-for-dashcard` it is difficult to properly set
`qp` for all of the dashcards without doing card lookups. So [this commit](https://github.com/metabase/metabase/blob/release-x.50.x/src/metabase/query_processor/card.clj#L230-L232) did a check to swap the runner in the presence of a pivot query that calls the **default** runner with a `qp.pivot/run-pivot-query`. The problem is that pulse needs to use a non-default runner, but there's no way for `process-query-for-card` to do both.

This commit changes `run` to `make-run` which raises the abstraction to
a function that produces a runer given a `qp`. This allows
`process-query-for-card` to change the `qp` as needed while keeping the
passed in runner.

There's really three things going on that could be teased apart
further/better.

1. `qp` - this conflates the processor (`qp/process-query` `qp.pivot/run-pivot-query`)
  and modifying query (`qp/userland-query`,
  `qp/userland-query-with-default-constraints` `(update query :info merge info)`)
2. `make-run` - How the above are processed: (streaming/realized)

* Fix tests

Tweak the callout about the SDK still being in development

* wip

* Remove references to the old enable-query-caching setting

* remove fixme

* linter

* remove some more instances

* fix imports

* skip super slow xray test

* unmark not very slow tests

* Elevate include this period in the relative date picker (#44179)

* Mock new row with "include {period}"  string

* Implement "include this" using a toggle

* Elevate "starting from" option in the relative date picker (#44242)

* Elevate "Starting from" option in the filter picker

* Tweak the date offset close icon

* Fix tests

* fix column scaling issues in waterfall chart

- refactor dataset scaling to behave like scatter and cartesian chart models

* add E2E test

* fix charts crash in safari when dataset timezone is not in iana format

* Update Loki Snapshots

* better fix: apply offset instead of finding a timezone with such offset

* remove unused specs

* Update Loki Snapshots

* review

---------

Co-authored-by: Metabase Automation <github-automation@metabase.com>

Fix fields from Amazon Athena Tables failing to sync if there is not an underscore in the Table Name (#44032)

* Fix a potential bug in `hydrate-can-restore`

I had a bug here. I wanted to hydrate `:collection` on the items and
then return the original items, but instead I was `dissoc`ing
`:collection` from the items. If the original items passed in already
had `:collection` keys this would break them.

AFAIK this didn't cause any issues, but it could have - so fixing it here.

* Add a test

* Present collection items in the collection

When fetching collection items, we need to fetch the true
`collection_id` in order to figure out things like `can_restore` and
`can_delete`, but set it to the displayed/effective `collection_id` when
we actually send it out.

This fixes a bug where collections had their `collection_id` match their
`id` when returned by this endpoint.

* replace color-mod with color-mix

* update alpha values to round to 5%

* add deprecation notice to lighten, darken and alpha

* update loki reference image

* add alpha value css variables to storybook preview

* update loki reference image

Previously we were just adding this when the item was archived. Instead,
always send a value - if the item is not archived, it can't be restored,
so it will be `false` in this case.

Making this change will make fixing a frontend bug easier, because the
frontend will be able to just trust the `can_restore` value coming in
from the backend.

fix(sdk): Update README to move `Embedding the collection browser` to `Features currently supposed` (#44208)

* add additional release slack message

* trigger slack messages from build

* fixes

* remove silly var name

* update release script to trigger ops issue creation

* update secret vars

* combine collections e2e specs

* obey the linter