Never report stale prices

[SupernaviX]

The oracle uses many APIs to get price data for the various tokens it reports on. If one of those APIs is down (or not pushing us updates for whatever reason), there's two things we can do about it:

1. keep using the most recent prices reported by that API
2. stop reporting prices from that API

The node was doing 1, to avoid our own downtime by using stale prices. This PR switches it to do 2, so that prolonged outages don't cause us to report extremely old prices.

The changes:

• Before, nodes would always have some price for every token; now, if a node can't get updates for that token, it won't produce any price feeds which depend on it.
• We no longer use "default" prices the first time an oracle starts up
• We won't ever reference prices saved to disk (though we still save them to disk in case we want to start doing that again someday)

[Quantumplation]

Do we have a metric for stale prices, or are we just relying on "we haven't reported prices for 2 minutes"?

Might make reporting easier if we had the former