Replace redux-form with react-hook-form

Implementation of the spike which recommends that we continue replacing redux-form with react-hook-form

In the linked spire there are a few PRs that have successfully replaced redux-form with react-hook-form, there's a number of loose ends to tidy up e.g. no longer using redux-form helpers to update the redux store

Acceptance criteria

[ ] redux-form dependency is removed entirely
[ ] Any references to redux-form, including code comments, are removed or updated as appropriate
[ ] redux store continues to be updated as needed, without redux-form helpers
[ ] Any API change to FormBuilder component is explicitly noted in the CMS 6 changelog and an upgrade path is provided
[ ] Kitchen-sink CI is run using all PRs and Behat enabled
[ ] Form UI remains unchanged.
[ ] yarn warning about providerValue in silverstripe/admin is resolved
[ ] Elemental TextCheckboxGroupField renders correctly
[ ] Elemental handleFormInit / onFormInit functionality works correctly

Notes

We may end up ditching redux altogether for form management, but not now.

PRs

POC PRs

In the PRs above I've swapped out redux-form for react-hook-form and run it through a CI run of kitchen-sink including all of the behat tests, and there were a fairly large number and wide range of failures. I had a decent look at one of the failures and it was because the AnchorSelectorField relies on the redux state set by redux-form of an entirely different react component in order to function.

Currently, the form state is updated by reducer provided by redux-form, which updates window.ss.store.getState().form.formState. This reducer is connected in registerReducers.js.

In order to fix all of the behat failures, we'd probably need to replicate what the redux-form reducer does, though seems like a pretty poor use of time. Not only that, this would be adding more complexity to an already convoluted tech stack and actually worsen the current situation.

A better approach would be to aim for removing Redux from our tech stack entirely.

This would change the approach taken here. Instead of starting by swapping redux-form with react-hook-form, we should keep redux-form temporarily and gradually refactor Redux out of individual React components, replacing it with regular component state. Once all components are refactored, the redux-form updates to Redux state become irrelevant. At that point, we can replace redux-form with react-hook-form.

Refactoring all of the react components that use redux is a massive increase in scope. While I think it's worthwhile long term to improve the code base, realistically it's hard to justify the opportunity cost.

A quick search in silverstripe for mapStateToProps in ./vendor/silverstripe/**/src/* yields 32 matches, so it's a large epic sized job.

Feature files that failed on kitchen-sink ci

admin

gridfield-search.feature
gridfield-with-arbitrarydata.feature
gridfield-with-arraydata.feature
manage-users.feature

asset-admin

file-details.feature
insert-an-image.feature
manage-files.feature
replace-file.feature

campaign-admin

manage-campaigns.feature

cms

insert-anchor-link.feature
search-for-a-page.feautre

elemental:

add-block-element-to-data-object.feature
add-link-to-anchor.feature
inline-block-validation.feature
page-save-validation.feature
publish-block-element.feature
restore-archived-element.feature
validation-failure.feature

elemental-bannerblock

block-link-field.feature
element-editor.feature

linkfield

accessibility-linkfield.feature
create-edit-linkfield.feature
linkfield-in-elemental.feature

silverstripe / silverstripe-admin