Report a parked block's last check and return its result

[nbenn]

Summary

• A parked block now reads the status its last check reached (ready, failed, waiting or unset) while nothing that check read has changed, stale once something has, and unevaluated without a check, which also covers a board block that is not built yet. The dormant status goes: it reported that nothing was computing the block rather than what a run would find, so a consumer could not tell a current parked block from one that failed, cannot run, was edited since or never ran. This is the first work item of https://github.com/orgs/BristolMyersSquibb/discussions/488 and replaces the direction this PR first took, which kept dormant with a narrower meaning.
• The check record from that first direction stays. Each check of a needed block, whether it runs the block or finds that it cannot, records the expression, state and eval trigger, which block feeds each data input and what each of those held, and now also the status it reached. A parked block compares against that record, an upstream counting as changed when it no longer holds what the block consumed from it or reads stale or unevaluated itself. An expression built from input data, such as blockr.dm's dm_select, cannot be rebuilt while its inputs are withheld, so for such a block the state it is built from is compared instead. The block server now owns the status reactive, status in its return list in place of failed, and board_server() installs it as the block's rv$eval slot.
• A parked block's result() returns what its last check left rather than NULL, so a block reading ready hands back the result that status describes (A block that is off screen returns NULL from result(), even after it has run #363). A block last found unable to run holds NULL. Output rendering follows the status for a parked block as well, which keeps the render observer from validating it against withheld inputs now that a parked upstream reads its held ready.
• The check, not the render observer, records why a block cannot run. A block checked off screen by an evaluate request now carries the status-phase reason, and one fixed off screen drops it once it runs. Before, the first read dormant with no conditions and the second kept its old note.
• Rebased onto main after Fold front-end demand into the one multi-owner claim set #337, with the tests holding blocks eager instead of required. The unreleased 0.1.4 NEWS entries that named dormant (Dormant downstream block's status stays stale after an upstream change (deferred eval surfaces no dirty signal) #310, Provide a channel to evaluate a dormant block without making it visible #318, Collapsed stacks evaluate once before gating engages #343) are corrected in place.

Migration

Before	After
A parked block reads dormant, whatever its last run found	It reads its last check's status while that is current, stale once it is not
A block that never ran reads dormant, and an unbuilt one has no status	Both read unevaluated
A parked block's result() is NULL	It is the result the block's last check left
An evaluate request waits while a block reads dormant or stale	It waits while a block reads unevaluated or stale, so a request for a current parked block is spent at once

Downstream, blockr.dock's block_status_badge() maps dormant to NA and so keeps the last badge; with this change it draws the held status instead, and BristolMyersSquibb/blockr.dock#485 drops the dormant case. In blockr.assistant, eval_deferred() and the status notes list dormant; reading the held statuses there is BristolMyersSquibb/blockr.assistant#165.

Fixes #362
Fixes #363

[codecov]

Codecov Report

❌ Patch coverage is 96.87500% with 3 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
R/block-server.R	96.55%	3 Missing ⚠️

Files with missing lines	Coverage Δ
R/block-class.R	96.16% <ø> (ø)
R/board-server.R	97.67% <100.00%> (+0.09%)	⬆️
R/plugin-code.R	100.00% <ø> (ø)
R/block-server.R	96.68% <96.55%> (+0.04%)	⬆️

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[nbenn]

Reworking this for https://github.com/orgs/BristolMyersSquibb/discussions/488. The check record stays, but a parked block now reports its last check's outcome while nothing it read has changed, stale once something has, and unevaluated without a check, so dormant goes instead of taking a narrower meaning. The reason a block cannot run moves into the check, and #363 is folded in, since a parked block reading ready has to return its held result. The title and description follow with the rework.