feat(web): optimization via lazy construction of OSK layers ⏩

[jahorton]

Continues from #11263, addressing part of #11166.

One particularly easy way to load and display the OSK more quickly is to simply... not make the parts you don't need yet. We can do just-in-time (JIT) construction of OSK layers when they're needed.

While the default layer is pretty much always needed, and some keyboards will sometimes start in one or two other layers... a large number of keyboards have extra layers that can wait.

As there's one more followup on this set of PRs...

@keymanapp-test-bot skip

[keymanapp-test-bot]

User Test Results

Test specification and instructions

User tests are not required

Test Artifacts

• Android
  • Keyman for Android apk
  • FirstVoices for Android 16.0 apk
  • FirstVoices Keyboards for Android apk
  • KeyboardHarness apk
  • Keyman for Android 16.0 apk
  • KMSample1 apk
  • KMSample2 apk
• Developer
  • Keyman Developer
  • Compiler Regression Tests
  • kmcomp.zip
• iOS
  • Keyman for iOS (simulator image)
  • FirstVoices Keyboards for iOS (simulator image)
  • TestFlight internal PR build version - 18.0.59 (0.11264.11327)
• Keyboards
  • Test Keyboards
• Web
  • KeymanWeb Test Home
• Windows
  • Keyman for Windows
  • FirstVoices Keyboards for Windows
  • Text Editor (32 bit)
  • Text Editor (64 bit)

[jahorton]

Current build-failure cause:

15:26:42   renderer_core.ts(181,54): error TS2341: Property 'layers' is private and only accessible within class 'OSKLayerGroup'.

Right, some of the tools will want access to that.

[ermshiperete]

This line seems unnecessary since that's already done in buildLayer.

[jahorton]

Build failure seen for "Test: Language Modeling Layer (Common)":

15:56:57 
  LMLayer using the trie model

15:56:57 
    Prediction

15:56:57 
    Regression tests

15:56:57 
      should use the default searchTermToKey()

15:56:58 
        expected +0 to be at least 1

15:56:58 
        AssertionError: expected +0 to be at least 1
            at file:///C:/BuildAgent/work/99b311828f4ee7c/keyman/common/predictive-text/unit_tests/headless/worker-trie-integration.js:85:16

The worker's "test mode" is on, so we shouldn't be exiting early... but perhaps recent changes from #11784 and its predecessors have had an unintended consequence here.

[keyman-server]

Changes in this pull request will be available for download in Keyman version 18.0.60-alpha

[mcdurdin]

Do you have any stats on improvement in performance?

[mcdurdin]

Why this change?

[jahorton]

Took a bit to find it, but the reason is on line 1319. It shifts from using the set of pre-built OSK layers (which have a .rows property) to using the underlying touch-layout specification (which uses .row, not .rows) as the basis for our calculations here.

[mcdurdin]

Understood. A bit squidgy, but make sense!

[jahorton]

Do you have any stats on improvement in performance?

Not in isolation. I did collect stats for the set of three it was a part of though, and that can be found on #11265. Note the last bit before the "User Testing" section in particular - that's the group stats.