[js/web] Forward WebGPU EP buffer cache mode options from JS

[ssam18]

Description

The native WebGPU EP already supports the buffer cache mode options (ep.webgpuexecutionprovider.storageBufferCacheMode and friends), but onnxruntime-web never forwarded them from executionProviders, so they were unreachable from JS. This adds storageBufferCacheMode, uniformBufferCacheMode, queryResolveBufferCacheMode and defaultBufferCacheMode to WebGpuExecutionProviderOption and forwards them to the EP the same way validationMode is forwarded today, with the values validated against the set the native side accepts. The options ride the existing SessionOptionsAppendExecutionProvider path, which prefixes each key into exactly the config entry the EP reads, so no native changes are needed.

Motivation and Context

Fixes #29016. For static shape models, storageBufferCacheMode: 'simple' reuses exact size buffers across runs instead of allocating new bucket sized ones, which the issue's repro shows cutting peak WebGPU memory by about 27 percent. Verified locally with tsc builds of js/common and js/web, prettier and eslint, the js/common unit tests, and type level checks that the new options compile and invalid values are rejected.

[popelenkow]

Thanks for the quick fix.

One concern: should defaultBufferCacheMode be exposed in this PR?

The native side declares/parses ep.webgpuexecutionprovider.defaultBufferCacheMode, but it does not appear to be wired into the active BufferManager.

It is parsed here together with the other buffer cache modes:

onnxruntime/onnxruntime/core/providers/webgpu/webgpu_provider_factory.cc

Lines 243 to 247 in cf509d8

	WebGpuBufferCacheConfig& buffer_cache_config = config.buffer_cache_config;
	parse_buffer_cache_mode(kStorageBufferCacheMode, buffer_cache_config.storage.mode);
	parse_buffer_cache_mode(kUniformBufferCacheMode, buffer_cache_config.uniform.mode);
	parse_buffer_cache_mode(kQueryResolveBufferCacheMode, buffer_cache_config.query_resolve.mode);
	parse_buffer_cache_mode(kDefaultBufferCacheMode, buffer_cache_config.default_entry.mode);

But when the main WebGPU BufferManager is created, only storage/uniform/queryResolve modes are passed:

onnxruntime/onnxruntime/core/providers/webgpu/webgpu_context.cc

Lines 148 to 151 in cf509d8

	buffer_mgr_ = BufferManagerFactory::Create(*this,
	config.buffer_cache_config.storage.mode,
	config.buffer_cache_config.uniform.mode,
	config.buffer_cache_config.query_resolve.mode);

BufferManager itself also accepts only those three modes, and default_cache_ is hardcoded to BufferCacheMode::Disabled:

onnxruntime/onnxruntime/core/providers/webgpu/buffer_manager.cc

Lines 490 to 495 in cf509d8

	BufferManager::BufferManager(WebGpuContext& context, BufferCacheMode storage_buffer_cache_mode, BufferCacheMode uniform_buffer_cache_mode, BufferCacheMode query_resolve_buffer_cache_mode)
	: context_{context},
	storage_cache_{CreateBufferCacheManager(storage_buffer_cache_mode)},
	uniform_cache_{CreateBufferCacheManager(uniform_buffer_cache_mode)},
	query_resolve_cache_{CreateBufferCacheManager(query_resolve_buffer_cache_mode)},
	default_cache_{CreateBufferCacheManager(BufferCacheMode::Disabled)} {

The factory also has no fourth/default cache mode parameter:

onnxruntime/onnxruntime/core/providers/webgpu/buffer_manager.cc

Lines 635 to 637 in cf509d8

	std::unique_ptr<BufferManager> BufferManagerFactory::Create(WebGpuContext& context, BufferCacheMode storage_buffer_cache_mode, BufferCacheMode uniform_buffer_cache_mode, BufferCacheMode query_resolve_buffer_cache_mode) {
	return std::make_unique<BufferManager>(context, storage_buffer_cache_mode, uniform_buffer_cache_mode, query_resolve_buffer_cache_mode);
	}

So exposing defaultBufferCacheMode from JS may make it look configurable even though it is effectively ignored unless the native BufferManager wiring is updated as well.