Expose KEDA ScaledObject and HTTP scaling configuration options

## Problem

Several KEDA scaling configuration values are currently hardcoded and not configurable via func.yaml.

### ScaledObject (Kafka trigger)

Hardcoded in `pkg/keda/kafka_scaling.go`:
- `cooldownPeriod`: 300 (seconds) — how long before scaling down after triggers go idle
- `pollingInterval`: not set, defaults to 30 seconds — how often KEDA checks metrics

Not exposed from KEDA's Kafka scaler:
- `offsetResetPolicy`
- `allowIdleConsumers`
- `excludePersistentLag`
- `scaleToZeroOnInvalidOffset`
- `limitToPartitionsWithLag`

### HTTPScaledObject (HTTP trigger)

Hardcoded in `pkg/keda/deployer.go`:
- `CooldownPeriod`: 300
- `ScalingMetric.Rate.TargetValue`: 100 (requests/sec)
- `ScalingMetric.Rate.Window`: 1 minute
- `ScalingMetric.Rate.Granularity`: 1 second

## Proposed solution

Add optional fields to `scale.keda` in func.yaml for the common options:

```yaml
deploy:
options:
scale:
keda:
pollingInterval: 15
cooldownPeriod: 60
triggers:
- type: kafka
lagThreshold: 5
- type: http
targetValue: 50
```

Start with `pollingInterval`, `cooldownPeriod`, and HTTP `targetValue` as the most impactful. The Kafka-specific options (`allowIdleConsumers`, etc.) can be added later as needed.

Start with pkg/keda/kafka_scaling.go and pkg/keda/deployer.go, then trace how scale.keda options from func.yaml are represented and passed into these KEDA configurations. Implement the proposed optional pollingInterval, cooldownPeriod, and HTTP targetValue fields first, and verify that configured values replace the current defaults without changing behavior when omitted.

Written by the indexing model from the issue text.