Fall back to per-target defaults for sparse Esplora fee estimates

[Jolah1]

Fixes #1028.

Esplora may return a non-empty estimate map that still has no usable entry for a given confirmation target, since convert_fee_rate only yields a value if the map holds a block count at or below the requested one. A map of {144:
2.0, 1008: 1.0} therefore leaves every target below 144 blocks — UrgentOnChainSweep, OnchainPayment, ChannelCloseMinimum, ... — without an estimate.

Today we substitute 1 sat/vb in that case. That's not a missing fallback but an actively harmful one: OnchainFeeEstimator::estimate_fee_rate already falls back to get_fallback_rate_for_target whenever it misses the cache, so
writing 1 sat/vb into the cache only serves to shadow the value we'd otherwise have used. For UrgentOnChainSweep that's 250 sats/kwu where the default is 5000 — a 20x underestimate on a target whose whole point is timeliness.

So here we fall back to the target's own default instead. The post-estimation adjustments are deliberately skipped for it: there's no estimate to adjust, and skipping them lands us on exactly the rate the cache-miss path would
have produced, which keeps the two paths consistent.

One open question for reviewers: whether that's the behaviour you want, or whether the fallback should go through apply_post_estimation_adjustments anyway. It only matters for MaximumFeeEstimate (8000 unadjusted vs. 11300
adjusted) and MinAllowedNonAnchorChannelRemoteFee (unchanged, as the subtraction clamps back to the floor). Happy to flip it — it's a one-line change plus one expected value in the test.

To make this testable without an Esplora server, the cache construction moves out of update_fee_rate_estimates into a pure build_fee_rate_cache. The new tests cover sparse maps, empty maps on and off Mainnet, and assert the
complete-map behaviour is unchanged.

I confirmed the sparse test fails against the previous behaviour (250 vs. 5000 sats/kwu for UrgentOnChainSweep, 250 vs. 5000 for OnchainPayment).

[ldk-reviews-bot]

I've assigned @tnull as a reviewer!
I'll wait for their review and will help manage the review process.
Once they submit their review, I'll check if a second reviewer would be helpful.