Refactor Log and Error Types and Handling

[AaronPlave]

___REQUIRES_AERIE_PR___="1777"

Backend PR NASA-AMMOS/plandev#1777 unifies every Java exception response to one shape under errors[].extensions. @Mythicaeda identified two UI issues in her testing:

1. The UI silently dropped extensions.type — every backend error was relabeled as the generic CAUGHT_ERROR.
2. Exceptions thrown during scheduling/simulation/constraints ended up in the Logs tab instead of the operation's own tab (Scheduling / Simulation / Constraints).

This PR fixes both and additionally refactors/consolidates the error/log pipeline. Previously the pipeline used three different typed error interfaces, three stores, and three catch helpers to mirror the backend's three-different-response-shape model. Now that the backend returns one shape, the UI can mirror this with a single store and two helpers (catchError and logMessage). Additionally, this PR does a rename pass to change BaseError -> ConsoleEntry and errors.ts store to console.ts to better align with the realities of what they actually represent now.

Visible UX Changes

As a result of this PR, exceptions that occur from Scheduling, Simulation, etc. will appear in their respective console tabs instead of the generic Logs tab.

Changes

Unified pipeline — one writable store consoleEntries: Writable<ConsoleEntry[]> + two helpers replace three writable stores, three catch helpers, and the old allLogs writable. catchError(category, message, error) dispatches via instanceof Error to handle thrown JS errors and backend ConsoleEntry payloads. logMessage(category, message, options?) for info/timing logs. Per-tab readable stores (schedulingErrors, simulationErrors, constraintErrors, modelErrors, allLogs, errorLogs, allProblems) are derived stores filtered by type on consoleEntries.

Parser fix — reqHasura now extracts extensions.type, preserves cause/trace separately (was overwriting trace with cause), and simulate uses reason.type instead of hardcoded CAUGHT_ERROR. reqWorkspace and reqWorkspaceMetadata now parse the raw WorkspaceFormattedError body (workspace uses USE_HASURA_FORMATTING=false, so no extensions wrapper) and throw it as a CompoundError — workspace callers get typed errors with full backend metadata flowing into the Workspace Console's Logs tab. ErrorTypes enum expanded to cover 27 new backend types.

Call-site routing — schedule/simulate/checkConstraints tag their outer catch (e) with their own category and clear stale entries at start. This routing is done so that backend exceptions land in the appropriate tabs (we cannot rely solely on the backend exception service field as it could be aerie_permissions when the user was running a scheduling request).

Renames (cosmetic) — BaseError → ConsoleEntry; src/types/errors.ts and src/stores/errors.ts → console.ts; simulationDatasetErrors/constraintRunErrors → simulationErrors/constraintErrors.

Verification

Unit Tests

• New src/stores/console.test.ts
• New src/utilities/requests.test.ts

Manual

Run against a backend on refactor/permissions-service-exceptions:

1. Drop the scheduling spec for an existing plan via Hasura console using:

mutation DropSchedulingSpec {
  delete_scheduling_specification(where: { plan_id: { _eq: YOUR_PLAN_ID } }) {
    affected_rows
  }
}

2. Run scheduling from the plan. Verify that the scheduling tab badge increments to reflect a new error. Verify that the logs tab does not. Expanding the scheduling error row should show the full type, service, data, trace from extensions.
3. Repeat the this spec dropping pattern for Simulate and Check Constraints.
4. Trigger a generic non-tab failure (e.g. a save error) and confirm it still routes to the Logs tab as before.

Future Work

• Route workspace errors into appropriate toast messages (ex: add a file that already exists at that path, toast message should indicate this to user instead of just the logs)