fix: Avoid redundant table column measurements on re-render

[TrevorBurnham]

This PR reduces unnecessary DOM measurements when a table with "sticky" features (stickyColumns or stickyHeader) re-renders.

ColumnWidthsProvider runs a useEffect that depends on its visibleColumns array. That effect calls updateColumnWidths(), which synchronizes sticky cells by reading layout from the DOM via getBoundingClientRect() (the sticky branch of getColumnStyles). In InternalTable, the arrays passed to the provider — visibleColumnDefinitions and the derived visibleColumnWidthsWithSelection / visibleColumnIdsWithSelection — are rebuilt as fresh arrays of fresh objects on every render, causing the provider's effect to re-fire. These unnecessary calculations contribute to input lag when users type in a sticky table's searchbox.

Memoize the derivations in InternalTable:

• visibleColumnDefinitions — keyed on columnDefinitions, columnDisplay, and visibleColumns.
• visibleColumnWidthsWithSelection / visibleColumnIdsWithSelection — keyed on hasSelection and visibleColumnDefinitions.

Re-sync sticky headers on content-driven resizes:

Memoizing visibleColumns means the provider's effect no longer re-fires on every render, so sticky header copies are no longer re-synced from the primary cells on unrelated re-renders. That's the point for fixed-width and resizable columns, but auto-layout columns (no explicit width) can change width when their content changes without any change to the column definitions (e.g. paginating or filtering). To keep the sticky header copies correct in that case, ColumnWidthsProvider now observes the primary cells with a ResizeObserver and re-syncs the copies when a primary cell resizes. The observer re-attaches when the set of visible columns changes, and no-ops in environments without ResizeObserver (e.g. SSR).

Benefits

Measured getBoundingClientRect calls per re-render with unchanged columns:

Configuration	Before	After
stickyColumns	3	0
stickyHeader	5	3

The remaining reads come from StickyHeader's own per-render synchronization, a separate mechanism this PR does not change (a candidate for a follow-up).

Non-sticky tables benefit in a smaller way from reduced array allocations.

How has this been tested?

Added unit tests in src/table/__tests__/columns-width.test.tsx:

• does not measure column widths when re-rendering with unchanged columns — re-renders with new items but identical columnDefinitions and asserts no getBoundingClientRect calls occur. Fails on main, passes with this change.
• does measure column widths when the columns change — a negative control that re-renders with an added column and asserts measurement still occurs.
• re-syncs the sticky header copy when a primary cell changes width without a column change — simulates an auto-layout cell growing and asserts the observed resize re-syncs the sticky copy.
• renders without a ResizeObserver (e.g. server-side rendering) — asserts the provider renders without throwing when ResizeObserver is absent.

---

By submitting this pull request, I confirm that you can use, modify, copy, and redistribute this contribution, under the terms of your choice.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 97.60%. Comparing base (82568c5) to head (c6a0d0a).

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #4609   +/-   ##
=======================================
  Coverage   97.60%   97.60%           
=======================================
  Files         951      951           
  Lines       30730    30742   +12     
  Branches    11277    11278    +1     
=======================================
+ Hits        29993    30005   +12     
  Misses        730      730           
  Partials        7        7           

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 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.

[ernst-dev]

Looks like there's a valid failing test relating to table measurements, please take a look at the failure

[TrevorBurnham]

Looks like there's a valid failing test relating to table measurements, please take a look at the failure

Fixed!

[ernst-dev]

Sorry Trevor, I know this is annoying but can we hit the codecov target for these changes?

[TrevorBurnham]

Sorry Trevor, I know this is annoying but can we hit the codecov target for these changes?

Done!