[ISSUE-703] Add Serialization tests for CPU neural simulations

[d-kamath]

Closes #703
Closes #598

Description

This PR introduces three test files to validate the correctness of the serialization process:

1. SerializeFullFileTest.cpp: Runs a full simulation (e.g., 10 epochs) and saves the serialized output.
2. SerializationFirstHalfTest.cpp: Runs the first half of the simulation (e.g., 5 epochs) and saves the serialized output.
3. SerializationSecondHalfTest.cpp: Runs the second half of the simulation (e.g., 5 epochs), using the serialized output from the first half, and compares it with the full simulation's serialized output.
4. Implemented a bash script (run_serial_test.sh) to automate the testing process.

Checklist (Mandatory for new features)

• Added Documentation
• Added Unit Tests

Testing (Mandatory for all changes)

• GPU Test: test-medium-connected.xml Passed
• GPU Test: test-large-long.xml Passed

[stiber]

Can’t the full length simulation results just be cached? Or is that too much like a recession test?

[d-kamath]

Can’t the full length simulation results just be cached? Or is that too much like a recession test?

It is possible, the cached result will need to be updated if any member variable is changed. Shall I cache the result?

[stiber]

Can’t the full length simulation results just be cached? Or is that too much like a recession test?

It is possible, the cached result will need to be updated if any member variable is changed. Shall I cache the result?

This is actually an interesting issue. In principle, we should check when changes to things like member variables represent a breaking change to serialization/deserialization, because simulator users may have saved serialization files they want to re-use and they need to know that won't be possible across certain updates.

Are we using Cereal's versioning capability? Maybe we need a separate test that compares a cached and a new serialization file — a test that fails if they have the same version but don't match (but passes if they have different versions). I don't think the versioning capability is anything more than embedded documentation, so it seems like it would need to be something that gets caught by a test like that.

[stiber]

OK, just make a new issue on additional ideas.