Problem: the lint pipeline (`./gradlew playwrightLint`) was not enforcing
much. tsconfig was missing `DOM.Iterable`, ESLint had no globals
declared for the browser context inside `page.evaluate()` callbacks, and
type-coverage was permissive enough that several `(window as any)` casts
were riding along without complaint. The deferred chrome flake fix
that landed in master via uPortal-Project#692 did not address these gates because the
previous PR predated them. Bringing the suite under a strict gate
surfaces real type/style issues that we want to catch in CI.

Goal: enable the strict lint+type-coverage gate, fix the issues it
surfaces, and add the only tests/code change that wasn't already in
master from the original uPortal-Project#691 — the `unhidePortletOptionsMenus`
helper.

Changes — config:
- tsconfig.json: add `DOM.Iterable` to `lib` so spread/iteration over
  NodeList works without casts
- eslint.config.js: declare browser globals (window, document,
  navigator, URL, Element, HTMLElement, NodeListOf, etc.) so
  no-undef stops flagging legitimate uses inside page.evaluate
  callbacks; the file is otherwise unchanged
- tests/uportal-pw.config.ts: add `trace: "retain-on-failure"` and
  `screenshot: "only-on-failure"` for CI debug

Changes — `unhidePortletOptionsMenus` helper added in `ux-general-utils.ts`:
- The XSL renders `.portlet-options-menu` with `class="hidden"` and a
  jQuery document.ready handler removes the class on portlets that
  have menu items. In headless Playwright runs the timing is variable
  enough that the click target may still be hidden when a test
  interacts with it. The helper force-removes the class as a
  deterministic prep step. Pairs with the existing `openDropdown`
  helper — that one drives the Bootstrap-side, this one the
  upstream jQuery-side that gates whether the toggle is rendered.

Changes — under-the-hood typing improvements driven by the new gate:
- ux-general-utils.ts: declare `window.bootstrap` global so
  `openDropdown` doesn't need a triple-cast inside `evaluate`
- announcements.spec.ts: declare `window.tinymce` / `window.tinyMCE`
  globals; use `window.tinymce?.activeEditor?.setContent(...)`
  instead of `(window as any).tinymce.activeEditor.setContent(...)`;
  drop one `waitForLoadState("networkidle")` flagged by the
  playwright plugin; numeric separators on the 10_000 timeouts
- simple-content.spec.ts: declare `window.CKEDITOR` global with
  proper interfaces; use `window.CKEDITOR` directly in evaluate
  callbacks; replace one `waitForLoadState("networkidle")` with an
  explicit `waitForURL(/render\.uP/)`; replace nested
  setData-as-Promise with a fire-and-forget setData (the cancel
  test only needs *something* staged)
- calendar.spec.ts, jasig-widgets.spec.ts, web-proxy.spec.ts:
  replace `waitForLoadState("networkidle")` with concrete
  visibility / text assertions on each portlet's distinguishing
  element; rename `e` → `error` in pageerror handlers
- bookmarks.spec.ts: regex case normalization (the `i` flag handles
  case so the literal can be lowercase)
- feedback.spec.ts: `submitBtn` → `submitButton`
- favorites.spec.ts: numeric separator on `10_000`
- api/utils/api-portlet-list-utils.ts: `.length).not.toEqual(0)`
  → `.toHaveLength(0)` per the playwright/prefer-to-have-length
  rule

Notes: ran `./gradlew playwrightLint` clean (0 errors, 19 warnings on
pre-existing pattern usages that the rules flag as suggestions). Ran
the full Playwright suite three times in a row at 117/117 against the
released versions to confirm none of the lint-driven refactors changed
test behavior. The unhide helper is exported but not yet called from
any spec — that's intentional, this PR adds the tool; tests that need
it can pull it in as their `openDropdown` calls reveal a hidden-class
race in CI.