fix(nimbus): distinguish errored metrics from not-yet-computed metrics on results

[jaredlockhart]

Because

• The results page branches on has_errors or not has_data, so a metric Jetstream failed on and a metric whose analysis window has not run yet render identically.
• Every weekly and overall outcome metric on a recently launched experiment hits the second state, so the page offers a support-contact banner for an analysis that is working correctly.

This commit

• Splits the two states in the results template so errors keep the amber warning treatment and the support button.
• Renders a blue informational state for metrics with no data yet, without a support button.
• Moves the results availability copy into NimbusUIConstants.
• Adds view tests covering both states.

Fixes #17155

[jaredlockhart]

Local renders on a seeded experiment whose results_data was hand-written so one KPI metric is in v3.errors and another has no valid point estimate.

Both states on one card, so the contrast is visible along with the collapsed header icons:

Error state, unchanged from before: amber banner, warning triangle, Contact Experimenter Support button. Header pill shows an amber triangle.

No-data state, new: blue alert-info, circled-i, "Not available yet", no support button. Header pill shows a blue info icon.