prescribe_radiative_gas_concentrations values are not actually applied and are zero

### What happened?

During investigation of a dycore read bug, claude-fable:5 flagged that
> prescribe_radiative_gas_concentrations_init sets CFC11 from your prescribed_cfc11_vmr = 653.45e-12, and it runs at phys_init (cam_comp:286) — before the first phys_timestep_init → physics_read_data (cam_comp:375).

This means that `physics_read_data` clobbers the contents of `prescribe_radiative_gas_concentrations_init`.

This needs to be investigated as it implies that the CAM4 suite as-is in CAM-SIMA is broken.

### What are the steps to reproduce the bug?

observe QPC4 regression test run log
```
Constituent cfc12 default value not configured. Setting to min value of
0.000000000000000E+000
Constituent o3 default value not configured. Setting to min value of
0.000000000000000E+000
Constituent ch4 default value not configured. Setting to min value of
0.000000000000000E+000
Constituent o2 default value not configured. Setting to min value of
0.000000000000000E+000
Constituent n2o default value not configured. Setting to min value of
0.000000000000000E+000
Constituent cfc11 default value not configured. Setting to min value of
0.000000000000000E+000
Constituent co2 default value not configured. Setting to min value of
0.000000000000000E+000
```

Golden moment from Claude:

### What CAM-SIMA hash were you using?

sima0_16_000

### What machine were you running CAM-SIMA on?

CISL machine (e.g. cheyenne)

### What compiler were you using?

Intel

### Path to a case directory, if applicable

/glade/derecho/scratch/hplin/aux_sima_intel_20260716081334/SMS_Ln9.mpasa120_mpasa120.QPC4.derecho_intel.cam-outfrq_analy_ic_cam4.GC.aux_sima_intel_20260716081334/run

### Will you be addressing this bug yourself?

Yes

### Extra info

_No response_

Start with prescribe_radiative_gas_concentrations_init at phys_init in cam_comp:286, then trace physics_read_data at cam_comp:375 and the QPC4 regression log. Determine why physics_read_data overwrites the prescribed values, and verify that the configured CFC and other gas concentrations are no longer reset to zero in the regression run.

Written by the indexing model from the issue text.