Issues after derivatives hooks are added

[wupeifan]

1. Currently we muted the gradient tests of the DSGE model solutions on CI. In particular, https://github.com/HighDimensionalEconLab/DifferentiableStateSpaceModels.jl/blob/main/test/first_order_sequence.jl#L42-L209 and https://github.com/HighDimensionalEconLab/DifferentiableStateSpaceModels.jl/blob/main/test/second_order_sequence.jl#L31-L143 are commented out. THESE TESTS WORK, but it takes an extremely long time for the CI (around 2 hours), while it works locally on my laptop with the attached manifest.toml file (have to zip it to upload to this GitHub issue)
Manifest.zip
2. Due to a bug with Zygote, the derivatives of the namedtuples are not well handled when Turing+Zygote are jointly used. After FluxML/Zygote.jl#1104 is merged from Zygote, we can uncomment the rbc_estimation test file. With that Zygote patch, our estimation MWE can work through.

[wupeifan]

A follow up: Most of the gradient tests don't take much time on CI. I unmuted most of them, and hence sequence.jl got tested.

[wupeifan]

Resolved with Zygote 6.30, and new version bumped to 0.4