fix(elements): include reasoning and cache costs in context total

[ephraimduncan]

The Context popover could show non-zero Reasoning and Cache line items directly above a "Total cost" that excluded both, so the visible breakdown didn't add up and users could underestimate a response's cost. The footer now prices the same four usage categories the row components display — input, output, reasoning tokens, and cached input reads — keeping all model-specific pricing in tokenlens rather than summing rounded strings.

Added a footer test using vercel:xai/grok-3-mini-fast (a model that prices all four categories) with token counts sized so omitting any single category changes the cent-rounded rendered total; it fails against the previous footer, passes now, and also asserts the unrounded aggregate reconciles with the sum of the four per-row prices.

[vercel]

@ephraimduncan is attempting to deploy a commit to the Vercel Team on Vercel.

A member of the Team first needs to authorize it.