Removes the Material import from the sliver tree rendering test by
replacing the MaterialApp wrapper with lower-layer test scaffolding that
preserves the fallback text metrics. Also removes the test from the
rendering cross-import allowlist.

Part of #177412
Refs #177028

Tests:
- `bin/cache/dart-sdk/bin/dart --enable-asserts
dev/bots/check_tests_cross_imports.dart`
- `bin/cache/dart-sdk/bin/dart analyze
packages/flutter/test/rendering/sliver_tree_test.dart
dev/bots/check_tests_cross_imports.dart`
- `./bin/flutter analyze --no-pub
packages/flutter/test/rendering/sliver_tree_test.dart
dev/bots/check_tests_cross_imports.dart`
- `./bin/flutter test --no-pub
packages/flutter/test/rendering/sliver_tree_test.dart`
- `git diff --check`

…... (#187279)

If this roll has caused a breakage, revert this CL and stop the roller
using the controls here:
https://autoroll.skia.org/r/fuchsia-linux-sdk-flutter
Please CC codefu@google.com,zra@google.com on the revert to ensure that
a human
is aware of the problem.

To file a bug in Flutter:
https://github.com/flutter/flutter/issues/new/choose

To report a problem with the AutoRoller itself, please file a bug:
https://issues.skia.org/issues/new?component=1389291&template=1850622

Documentation for the AutoRoller is here:
https://skia.googlesource.com/buildbot/+doc/main/autoroll/README.md

…ynamic boxing churn under Wasm (#186978)

Avoids heap allocations and GC sweeps caused by dynamic boxing of
primitive integers (struct allocations) inside high-frequency paragraph
layout, segmenter, and line-breaking loops.

1. **Optimized Local Typed List Copy Loops**:
- Replaced generic closure parameters and generic `List<int>.generate`
copy loops inside `paragraph.dart` and `path.dart` with specialized
local pointer extensions (`toUint8List` / `toUint32List`).
- Standard FFI `Pointer.asTypedList` is not supported at runtime under
`dart2wasm` in the Web Engine's external memory layout (since FFI
allocations are mapped to Skwasm's C++ heap rather than the standard
Dart heap, leaving the global FFI helper null at runtime).
- While these custom extensions still perform a memory copy step, they
allocate concrete, specialized typed arrays (`Uint8List` / `Uint32List`)
and copy elements using unboxed, register-level pointer lookups.
- Loops are structured to iterate backwards (`length - 1` down to `0`).
Compiling this to WebAssembly translates the loop condition to a direct
comparison against a constant zero register (`i64.ge_s` to `i64.const
0`), eliminating a register load instruction (`local.get $length`)
inside the high-frequency loop body on every iteration.
- Under `dart2wasm`, the copy loops are fully inlined and compile to raw
WebAssembly array loads and stores (`i32.load` / `i32.store`) with **0
generic closure callbacks** and **0 dynamic integer boxing** ($BoxedInt
heap allocations).
- For signed `Pointer<Int8>` buffers (such as in `path.dart`'s
`toSvgString`), the copy block explicitly masks values with `& 0xFF` to
convert them to positive unsigned bytes ($0$ to $255$), preventing
runtime `FormatException` issues under `utf8.decode` at zero memory
overhead.

2. **Pattern-Based Switch Expression Comparisons**:
- Replaced generic `Set<int>` lookup collections (`_kNewlines`,
`_kSpaces`) inside `breakLinesUsingV8BreakIterator` with
pattern-matching switch expressions (`_isNewline`, `_isSpace`).
- Under `dart2wasm`, this completely eliminates the dynamic boxing of
checked primitive `codeUnit` integers into heap-allocated `$BoxedInt`
wrappers (originally required for generic `Set.contains` lookup
parameters).
- Because the Unicode newline case numbers are sparse, the compiler
lowers the switch statement to inline, register-level direct comparisons
and conditional jumps (`i64.eq` and `br_if`) rather than an indirect
jump table (`br_table`), avoiding binary size bloat while executing
entirely on CPU registers with **0 heap allocations**.

Fixes #186972