Add Gemini 2.0 Flash and 2.5 Pro pricing to the model catalog

[navintkr]

The router can only recommend models that exist in
src/data/pricing.ts. The current catalog has
gemini-flash but is missing Gemini 2.0 Flash, 2.0 Flash-Lite, and
2.5 Pro, which are the variants most teams actually have access to
via vscode.lm today.

What to do

1. Open src/data/pricing.ts.
2. Look at the existing entries — each model is an object with
   id, tier, inputPricePerM, outputPricePerM,
   premiumMultiplier, contextWindow, and goodAt tags.
3. Add entries for:
   • gemini-2.0-flash (cheap tier, 0× premium)
   • gemini-2.0-flash-lite (cheap tier, 0× premium)
   • gemini-2.5-pro (reasoning tier, check Google's current multiplier)
4. Pull current prices from https://ai.google.dev/pricing — use the
   paid-tier numbers, not the free-tier ones.
5. Make sure the id matches the exact string vscode.lm.selectChatModels
   returns for that family (check by running the chat participant
   against a workspace that has Gemini enabled).

Acceptance

• New entries present in MODEL_CATALOG.
• @proctor /explain gemini-friendly prompt lists them in the
  candidate matrix.
• README model list updated if you count from it.

Why this matters

Without these entries, the router will never recommend them even
when Gemini is the cheapest model available, silently pushing users
toward OpenAI or Anthropic by default.