fix(processors): forward inputs through windowed sketch processors

[zzylol]

Summary

Stacked on the merged #210. Makes a single pipeline like `[ddsketch, HLL, batch]` actually work — without this, chained windowed sketch processors silently drop everything.

Bug

`ddsketch / kll / hll` processors in `mode: window` returned `nil` from `ConsumeMetrics` — they accumulated inputs into their windowed state and dropped the input batch on the floor without forwarding it to the next consumer:

```go
case ModeWindow:
p.accumulateIntoWindow(md)
return nil // <-- inputs go nowhere
```

In a chained pipeline:

1. otlp delivers raw → `ddsketch.ConsumeMetrics`. Accumulate. Return nil. HLL never sees the raw input.
2. ddsketch tick fires → emits typed-DDSketch → `HLL.ConsumeMetrics`.
3. `HLL.accumulateIntoWindow`'s switch only handles `Gauge` / `HLLSketch` — DDSketch falls through and HLL silently drops.
4. batch sees nothing.

`countminsketchprocessor` / `countsketchprocessor` already do the right thing — their ModeWindow paths return `md` (forward inputs) by default, with an opt-in `drop_original` flag. The other three were the outliers.

Fix

Match the CMS/CountSketch pattern in all three processors:

```go
case ModeWindow:
p.accumulateIntoWindow(md)
return p.nextConsumer.ConsumeMetrics(ctx, md)
```

Now the chain works:

• `ddsketch.ConsumeMetrics(raw)`: accumulate, forward raw to HLL.
• `HLL.ConsumeMetrics(raw)`: accumulate, forward raw to batch.
• `batch.ConsumeMetrics(raw)`: batch & ship.
• ddsketch tick → typed-DDSketch → HLL (no match, forward) → batch.
• HLL tick → typed-HLL → batch.

Both sketches reach the exporter alongside raw inputs. Users who don't want raw inputs shipped can add a `filter` processor at the end (or use `drop_original` on CMS/CountSketch).

Verification

End-to-end with the original chained `b3-delta` config (`processors: [ddsketch, HLL, batch]`, `delta_transmission: true`, `window_duration: 60s`):

```
W1 (16:21:09):
Ddsketch received: dps=1000
Hllsketch received: dps=1000 ← HLL now reaches backend
ingest: 1000 routed, 0 decode-failed, 1000 unconfigured

W2 (16:22:09):
Ddsketch received: dps=1000 (delta — applied via cached snapshot)
ingest: 1000 routed, 0 decode-failed, 1000 unconfigured
```

Supersedes #209

PR #209 worked around this bug at the config level by splitting `[ddsketch, HLL, batch]` into two parallel pipelines. That goes against the user's stated goal of one pipeline that supports any controller-chosen combination of sketches. With this fix, the original chained config works directly. #209 should be closed.

Test plan

• DDSketch + HLL outputs both reach the backend through a single `[ddsketch, HLL, batch]` pipeline.
• Delta path still works on second window (depends on fix(ddsketch): make delta_transmission actually work end-to-end #210, now merged).
• Chain with all 5 sketches in one pipeline — pattern is identical, should work; not tested in this round but no reason for it not to.

🤖 Generated with Claude Code