fix: make :class col more generic to X type

[MilesCranmer]

@ablaom this should fix the issue you noticed with rowtable(X) types.

However this will currently result in type instability for NamedTuple input since I have to collect the columns due to this issue: JuliaAI/MLJModelInterface.jl#214. But it does fix the issue for other types.

[github-actions]

Benchmark Results

	master	1840f53...	master/1840f531077584...
search/multithreading	17.3 ± 1.4 s	17.4 ± 0.66 s	0.996
search/serial	27.7 ± 0.53 s	28.6 ± 0.87 s	0.969
utils/best_of_sample	1.7 ± 0.92 μs	1.72 ± 0.55 μs	0.988
utils/check_constraints_x10	11.2 ± 2.8 μs	11.2 ± 3 μs	0.994
utils/compute_complexity_x10/Float64	2.08 ± 0.11 μs	2.07 ± 0.12 μs	1
utils/compute_complexity_x10/Int64	2.04 ± 0.12 μs	2.01 ± 0.11 μs	1.01
utils/compute_complexity_x10/nothing	1.45 ± 0.13 μs	1.43 ± 0.13 μs	1.01
utils/insert_random_op_x10	5.89 ± 1.9 μs	5.78 ± 1.9 μs	1.02
utils/next_generation_x100	0.421 ± 0.099 ms	0.353 ± 0.098 ms	1.19
utils/optimize_constants_x10	0.0339 ± 0.0087 s	0.0333 ± 0.0075 s	1.02
utils/randomly_rotate_tree_x10	5.2 ± 0.59 μs	5.19 ± 0.58 μs	1
time_to_load	1.84 ± 0.0064 s	1.86 ± 0.004 s	0.993

Benchmark Plots

A plot of the benchmark results have been uploaded as an artifact to the workflow run for this PR.
Go to "Actions"->"Benchmark a pull request"->[the most recent run]->"Artifacts" (at the bottom).

[coveralls]

Pull Request Test Coverage Report for Build 12459338137

Details

• 13 of 13 (100.0%) changed or added relevant lines in 3 files are covered.
• 1 unchanged line in 1 file lost coverage.
• Overall coverage increased (+0.03%) to 94.786%

---

Files with Coverage Reduction	New Missed Lines	%
src/SymbolicRegression.jl	1	98.3%

Totals
Change from base Build 12441469062:	0.03%
Covered Lines:	3145
Relevant Lines:	3318

---

💛 - Coveralls

[ablaom]

I've asked someone to take a look at the MLJModelInterface issue, which doesn't look too difficult to address. I'll leave it up to you whether you proceed with this PR or not. This isn't urgent from my point of view. If you would like me test if this resolves the issue at #392 then I'll need a PR combining both changes please (eg. merge this and rebase #392). Ping me when you 're ready.