chore: migrate to new column-access mechanism (SDK 1.81.x)

.changeset/migrate-column-access-mechanism.md

@@ -0,0 +1,17 @@
+---
+"@platforma-open/milaboratories.top-antibodies.model": patch
+"@platforma-open/milaboratories.top-antibodies.ui": patch
+"@platforma-open/milaboratories.top-antibodies.workflow": patch
+---
+
+Update `@platforma-sdk` to 1.81.x and migrate the block model to the new column-access mechanism.
+
+The removed `ColumnCollectionBuilder` / `ColumnMatch` / `ArrayColumnProvider` APIs are replaced by the host-driven `ColumnsCollection` + `ColumnRecipe` API: column discovery and filtering run host-side via `discover`/`filter` selectors, and specs are fetched on demand through `getSpec()` instead of being pulled eagerly into the sandbox. The main table's `createPlDataTableV3` display rules move from `(spec) => boolean` lambdas to host-side `ColumnSelector`s. No user-facing behavior change intended.
+
+Follow-ups for 1.81: the renamed predicates (`isColumnLazy` → `isDataColumn`), the `LinkerParts` shape now passed to the `linker` label formatter, and `umapPf` / `umapPcols` moving off the deprecated `ctx.resultPool.getAnchoredPColumns` + `TreeNodeAccessor.getPColumns()` onto anchored `ColumnsCollection` discovery. Every `PlRef` → spec read goes through `getSpecByRef`, which absorbs the `ColumnAbsentError` that `Column(ref)` throws where `ctx.resultPool.getPColumnSpecByRef` used to return `undefined`.
+
+The filter/ranking wire format stays a leaf `PObjectId`: `matchToColumnId` reduces the discovered recipe id via `extractPObjectId`, and `dedupByLeafId` collapses reachability variants to one entry per column the way the old `findColumns` did. `bundleBuilder.addSingle` resolves a global `PObjectId` by ref and has no branch for the `ColumnDiscoveredId` a linker-reached hit carries, so persisted selections and the workflow keep working unchanged. Putting the full `ColumnUniversalId` on the wire needs an `__isDiscovered` branch in `addSingle` and in `processColumnId` first.
+
+Drop the dead `anchorName` field from `ScopedColumnId`, along with the unused `ScopedColumn` / `Column` types. `anchorName` was always the literal `"main"` and only mattered while the wire value was an anchored query id; nothing on the workflow side reads it, and the two UI lookups that filtered on it were selecting the first option in a list where every entry carried the same value. **This changes the shape of `BlockArgs`, so existing projects re-run the workflow once.** Stored `BlockData` keeps the field as an inert leftover — no migration needed.
+
+`getSpecByRef` guards on the column's resolution status instead of catching `ColumnAbsentError`, keeping exceptions out of the control flow. A missing column stays silently ignored, as before: the block works with whatever the upstream graph provides, and the `absent` status occurs in normal configurations.

.github/workflows/build.yaml

@@ -39,6 +39,11 @@ jobs:
       package-path: 'block'
       create-tag: 'true'
 
+      # Require the published `block` package to be bumped by a changeset on
+      # PRs (empty changeset or the `skip-changelog` label waives it). Needs
+      # the input to exist on the pinned `@v4` reusable workflow.
+      require-package-path-bump: true
+
       npmrc-config: |
         {
           "registries": {