fix: don't let a single unrenderable input abort the whole workflow render

[timurbazhirov]

Summary

• ExecutionUnitInput.render() rethrew on any Nunjucks template error. That propagates uncaught through ExecutionUnit.render() -> Subworkflow.render() -> Workflow.render() -> job.render(), which job-designer's JOB_WORKFLOW_SYNC/JOB_UPDATE reducers call unconditionally.
• Selecting a multi-material workflow (e.g. Valence Band Offset, whose interface-left/right units are keyed by MATERIAL_INDEX "1"/"2" into input.perMaterial) while the job still only has its single default material makes perMaterial[1]/[2] undefined, which the sprintf filter throws on - crashing the reducer and silently aborting the workflow selection itself. job-designer's onSelectWorkflowsSubmit only console.errors the rejected promise, so the UI just stays on the old workflow with no visible error to the user.
• Fix: fall back to the raw (unrendered) template on error instead of rethrowing, matching the function's existing this.rendered = rendered || this.template.content fallback for empty output. The debug console.error logging is preserved. The next job/material update re-runs render() and produces a correct preview once enough materials are assigned.

Root cause chain (for reviewers)

1. reference/job-designer/src/reducers/renderJobForDesignerState.js calls job.render() unconditionally on every JOB_UPDATE/JOB_WORKFLOW_SYNC.
2. Workflow.render() -> Subworkflow.render() -> ExecutionUnit.render() render every unit in every subworkflow, not just the one currently visible.
3. VBO's standata definition (reference/standata/data/workflows/workflows/espresso/valence_band_offset.json) has "Set Material Index" assignment units setting MATERIAL_INDEX to "0"/"1"/"2" for its combined/left/right branches, and several unit templates guard on subworkflowContext.MATERIAL_INDEX to index into input.perMaterial[...].
4. Right after selecting this workflow (before the user assigns the 3 required materials), the job still has 1 material, so perMaterial has only index 0 - perMaterial[1]/perMaterial[2] are undefined.
5. ExecutionUnitInput.render() calls the custom sprintf filter (reference/standata/src/js/utils/template.ts) with undefined, which throws TypeError: [sprintf] expecting number but found undefined - previously uncaught here, now handled.

Test plan

• npm test in reference/wode - 47/47 passing, no regressions.
• Live repro in a real browser session against a local web-app + Meteor server: before the fix, selecting "Valence Band Offset (2D)" on a job with the default single material silently kept the old "Total Energy" workflow active (confirmed via DOM inspection + matching console errors). After the fix (verified via a locally patched node_modules copy + full rspack cache clear/restart), the workflow correctly switches, all 7 VBO subworkflows populate, and the "average ESP" unit (the element job-espresso-valence-band-offset.feature needs) is present and clickable.
• Re-ran cypress run --spec cypress/e2e/jobs/run/job-espresso-valence-band-offset.feature in web-app: previously failed at "I open 'average ESP' subworkflow unit" (never found); now progresses past that entirely into a later, unrelated bug (duplicate "cutoffs" panel in Important Settings causing an ambiguous cy.focus() match - tracked separately, not part of this PR).

🤖 Generated with Claude Code

[VsevolodX]

LGTM