Sum sensor group updating with ignore non numeric=false

### The problem

I have a sensor group helper which aggregates energy measurement from the A and B phase of my cooktop.
When either one becomes unavailable, the sensor group temporarily falls back to the value of the other sensor, causing dips in the measurement. When the sensor comes back online and the sum increases back to the normal state, the sensor group statistics note an increase equal to the sensor coming back online. Due to both source sensors being of state_class=total_increasing, the sensor group is total_increasing as well.
This causes some wild outliers in statistics, messing up the energy dashboard.

Looking at the description for 'ignore non-numeric', I would expect the unavailable states to be ignored, and the sum only resumed when both sensors report a valid value.

The logs do not show 'unable to use state' for this sensor group.

### What version of Home Assistant Core has the issue?

core-2026.9.2

### What was the last working version of Home Assistant Core?

_No response_

### What type of installation are you running?

Home Assistant OS

### Integration causing the issue

group

### Link to integration documentation on our website

https://www.home-assistant.io/integrations/group/#sensor-number-and-input_number-groups

### Diagnostics information

_No response_

### Example YAML snippet

```yaml

```

### Anything in the logs that might be useful for us?

```txt

```

### Additional information

The source sensors are two phases of a Shelly 3EM (gen1). Its statistics do not show the anomaly.

Start with the group integration's sensor-number and input_number group documentation and reproduce the behavior using two total_increasing source sensors, one becoming unavailable. Compare the resulting group state and statistics with the documented ignore-non-numeric behavior. Done means unavailable values do not create a false increase when a source returns.

Written by the indexing model from the issue text.