Example shiny app as a sandbox

[jonocarroll]

This PR adds a shiny app in inst/shiny/sandbox which can load any of fdge, fpca, or ffsea as modules.

This is a barebones app with no additional features, merely a prototype for a sandbox in which modules could be loaded. This helped identify some minor issues (e.g. facilebio/FacileAnalysis#49 and something else happening with ffsea which crashes here). This also has some best-practices in terms of cleaning up destroyed modules. I've also used it to prototype the idea of keeping a running list of analyses performed - again, this is just a simple implementation.

Dataset loading is limited to FacileData:::exampleFacileDataSet().

[jonocarroll]

Since it's always easier to show than tell:

Screen.Recording.2022-05-05.at.11.11.22.am.mov

[jonocarroll]

Still TODO:

• properly namespace and destroy the filteredReactiveFacileDataStoreUI()
• actually store the results so they can be presented / interacted with
• ?

[jonocarroll]

@lianos this may be where I'm going wrong - I seem to get the fallback list(). Not doing this step produces something that seems to error. Any guidance appreciated.

[jonocarroll]

Now with filtering as an "output" that can be fed back in as an "input"

Screen.Recording.2022-05-11.at.4.43.20.pm.mov

Still TODO:

• properly namespace (and destroy) the filteredReactiveFacileDataStoreUI() - this may be the reason I'm getting multiple filters (some inaccessible) with the 'add' button but I'll dig further. It may be a bug in the UI, may just be my implementation
• actually store the results so they can be presented / interacted with
• store a provenance chain linking all outputs to their inputs, recursively
• present a network graph of the inputs and outputs - someone may add a filter then pca vs raw with dge then fsea
• find out why debug=TRUE is a necessary condition - filtering UI doesn't show without this
  • it's expected; table only shows when debug=TRUE https://github.com/facilebio/FacileShine/blob/73a67ae0d5733d19dcabfc3f2e678e4ec1829161/R/sampleFilter-module.R#L86
• limit "output" to only the methods compatible with the selected "input". Somewhat chicken/egg if it should also work as selecting "inputs" which are compatible with "outputs".
• ?

[jonocarroll]

This logic is growing a bit too big, but I think this is the correct way to extract out the relevant parts - feel free to correct me, I'm still learning how these parts work.

[jonocarroll]

The doubled filters issue is now resolved and the right number of samples propagates through the filter into the result and retrieval by a later analysis. The filteredReactiveFacileDataStore server logic invoked as part of ReactiveFacileDataStore isn't being destroyed (because it's not an observer in the top scope) so invoking a second filter loads the settings from the first - this could be dealt with via a more consistent module for filtering.