feat: add Fireworks AI as a supported model provider

[Sayt-0]

Summary

Adds first-class support for Fireworks AI as an OpenAI-compatible model provider via the alias map, following the existing groq/deepseek pattern.

Closes #3349

Definition of Done

Item	Status	Where
Alias entry fireworks (openai, https://api.fireworks.ai/inference/v1, FIREWORKS_API_KEY)	done	pkg/model/provider/aliases.go
Unit test asserts alias resolves and IsCatalogProvider("fireworks") is true	done	TestFireworksAlias
Provider appears in catalog via CatalogProviders()	yes (has a BaseURL)	aliases.go
task test and task lint pass	yes (see Validation)
Example YAML config using a Fireworks AI model	done	examples/fireworks.yaml

Beyond the stated DoD, this also wires auto-detection, a default model, the JSON schema, and the provider docs, mirroring how DeepSeek was added.

Design decisions

Decision	Rationale
Added to openModelHostProviders (system-message coalescing)	Fireworks fronts open-weight models (Kimi, Qwen, GLM, Llama, gpt-oss) whose chat templates may reject more than one leading system message. Consecutive system messages are coalesced, per issue #3344. First-party APIs with a fixed lineup (mistral, xai, deepseek) are deliberately excluded.
Listed in modelsDevAbsentProviders (example test)	models.dev catalogs Fireworks under the provider id fireworks-ai, not fireworks, so the example-validation lookup is skipped, the same treatment as ovhcloud.
Auto-detection priority between groq and deepseek	Keeps the open-model-host ordering consistent; precedence assertions updated accordingly.
Default model accounts/fireworks/models/kimi-k2-instruct	Fireworks headline open model; ids use the accounts/fireworks/models/<name> form.

Tests

Test	Coverage
TestFireworksAlias	Alias resolution and catalog membership.
TestFireworksProvider_EndToEndRequest	Real HTTP request through the full stack against a local OpenAI-compatible server: asserts POST, Bearer FIREWORKS_API_KEY, /chat/completions route, verbatim model, stream reassembly, and that consecutive system messages are coalesced into one.
TestFireworksLiveAPI	Live API call, skipped unless FIREWORKS_API_KEY is set, so the default run stays hermetic.
TestShouldMergeConsecutiveMessages_Gating	Asserts coalescing is enabled for fireworks.
auto_test.go cases	Detection, default model, precedence chain, integration with DefaultModels.

Validation

Gate	Result
task build	passes
golangci-lint run (whole repo)	0 issues
custom cops (go run ./lint .)	no offenses
go test ./...	touched packages pass; remaining failures are pre-existing SSRF/network tests unrelated to this change

Residual design uncertainty

The default model is accounts/fireworks/models/kimi-k2-instruct. If a different default is preferred (for example a Llama or Qwen id), it can always be changed.

[aheritier]

👋 This PR has merge conflicts with the base branch. Please rebase or merge the latest base branch and resolve them. I've moved it to draft and added status/needs-rebase; it'll be picked back up automatically once the conflicts are cleared.