HDF5 export fails with dictionnary type serialization (export result)

[ThomasMalletCodra]

Describe the bug
HDF5 export fails with dictionnary type serialization (export result)
After processing the signal with "Extract pulse feature", attempting to save the file in HDF5 format fails : "cannot serialize 'step' of type"

To Reproduce

1. Open DataLab
2. Create a signal
3. Run Computing → Extract pulse feature on the signal
4. Attempt to save the project in HDF5 format
5. Observe the save error

Expected behavior

• HDF5 export should succeed without errors after running and computing function ("Extract pulse feature" for example)
• Feature export should be able to serialize and export dictionnary type (for generic result data support)

HDF5 save error:

See related issue in DataLab : DataLab-Platform/DataLab#300

[dappham-CODRA]

The proposed approach is to make TableResult the serialization boundary and ensure that its exported representation only contains HDF5-safe, JSON-friendly values.

Concretely, attrs should be sanitized before export so that display-only Python objects, especially callables, are not persisted. This avoids leaking presentation logic into the stored result payload and prevents HDF5 export failures caused by non-serializable values.

In the same spirit, values produced by the computation layer should be normalized to plain scalar types before they are stored in the table. If a result is not truly numeric, it should be converted to a plain string representation rather than keeping richer Python objects such as enum-like values.

If some display intent still needs to survive persistence, it should be carried through simple serializable metadata, for example column format specifiers, instead of Python formatting functions. This keeps the persisted payload stable while preserving a clean separation between computation data and presentation