feat: add Perplexity as a first-class API provider

[james-pplx]

Summary

Adds Perplexity as a first-class API provider in Roo Code, targeting Perplexity's OpenAI-compatible chat-completions endpoint at https://api.perplexity.ai.

Models exposed (all 128k context):

• sonar
• sonar-pro (default)
• sonar-reasoning
• sonar-reasoning-pro

API key resolution order:

1. perplexityApiKey from settings
2. PERPLEXITY_API_KEY env var
3. PPLX_API_KEY env var (fallback)

Changes

• Types (packages/types/src/providers/perplexity.ts): PerplexityModelId, perplexityModels, perplexityDefaultModelId.
• Provider settings (packages/types/src/provider-settings.ts): perplexity added to the ProviderName union, perplexitySchema added to the discriminated union and merged schema, registered in modelIdKeysByProvider and MODELS_BY_PROVIDER.
• Secret keys (packages/types/src/global-settings.ts): perplexityApiKey added to SECRET_STATE_KEYS.
• Handler (src/api/providers/perplexity.ts): PerplexityHandler extending BaseOpenAiCompatibleProvider, with resolvePerplexityApiKey helper for env-var fallback.
• Factory (src/api/index.ts): wired into createHandlerForProvider.
• Profile validation (src/shared/ProfileValidator.ts): perplexity resolves to apiModelId.
• Webview UI (webview-ui/src/components/settings/providers/Perplexity.tsx): API-key field + sign-up link, mirroring Fireworks.tsx. Wired into ApiOptions.tsx, constants.ts, providerModelConfig.ts, useSelectedModel.ts, and validate.ts.
• Locales (webview-ui/src/i18n/locales/en/settings.json): perplexityApiKey / getPerplexityApiKey strings added.

Testing

src/api/providers/__tests__/perplexity.spec.ts covers:

• Correct base URL (https://api.perplexity.ai).
• API key from settings.
• PERPLEXITY_API_KEY env-var fallback.
• PPLX_API_KEY env-var secondary fallback.
• Settings key takes precedence over env vars.
• Throws when no key is configured.
• resolvePerplexityApiKey returns undefined for empty/unset cases.
• Default model is sonar-pro.
• Selecting sonar-reasoning-pro returns its ModelInfo.
• Unknown model id falls back to default.
• All four models exposed with 128k context.
• createMessage streams text content.
• createMessage passes the configured model id to the upstream client.
• createMessage propagates upstream errors.

Run with: npx vitest run api/providers/__tests__/perplexity.spec.ts

Test plan

• CI runs perplexity.spec.ts and passes.
• Manual: set perplexityApiKey in settings, select a Perplexity model, send a message, verify streaming response.
• Manual: unset settings key, set PERPLEXITY_API_KEY env, restart extension, verify provider works via env fallback.