Add ComponentArrays extension

[ErikQQY]

Replace: #345
Fix: #344

So in #344, the reason why ComponentArray failed with DenseEuclideanMetric is that the dense matrix in DenseEuclideanMetric doesn't have the same axes as the input, and when we do linear algebra operations on two arrays with different axes, the result is just an ordinary Julia array, for example:

julia> r_init = ComponentArray(; a=randn(5), b=randn(5))
ComponentVector{Float64}(a = [-0.8058038744185181, -0.08413831281388894, 0.7453256076416268, -0.03469137039683826, -0.2517275575798834], b = [-1.5630129100605425, 1.7031219617336555, -1.0997337141360415, 0.4409537319539482, 0.19944559484881833])

julia> getaxes(r_init)
(Axis(a = ViewAxis(1:5, Shaped1DAxis((5,))), b = ViewAxis(6:10, Shaped1DAxis((5,)))),)

julia> M = rand(10, 10);

julia> axes(M)
(Base.OneTo(10), Base.OneTo(10))

julia> M * r_init # clearly the axes of M and r_init don't match
10-element Vector{Float64}:
 -0.06957219975578224
 -0.24499462421604584
  0.26535313109443504
 -0.05499884350331631
 -0.11871623121978324
 -1.132138726898246
 -0.6221912360430668
 -0.7051768915767198
  1.3780763510373664
 -0.9263200146682737

But it's basically user's obligation to ensure consistent input data structure, so this PR ensures the input from user are axes-matching.

Some notes:

• See the tests for example uses of ComponentArrays.
• The code changes in src/adaptation/massmatrix.jl are not related to fixing ComponentArrays.

[yebai]

I'm not familiar with ComponentArrays. Can @mhauru, @penelopeysm, or @sunxd3 take a look, please?

[yebai]

Excellent work, @ErikQQY!

I only checked the main code (/src/) but didn't look at the tests carefully.

[yebai]

@ErikQQY, can you update the PR summary to explain what the issue was preventing the use of ComponentArrays and how this PR fix it? It is not self-evident from the code changes.

[yebai]

@penelopeysm, can you also review this please?

[penelopeysm]

Very happy apart from the single comment!

[yebai]

Excellent work @ErikQQY!