Higher order gradients in CellValues

[lijas]

New version of #891 so that I can see preview of docs. I also needed to rebase on master

Implementation of higher order gradients in CellValues. This was quite easy to do thanks to Knuts new FunctionValues and GeometryMapping.

• Tests for FaceValues
• Update changelog
• Type stable CellValues constructor (?)

[KnutAM]

Just posting idea from Slack here for a type-stable ValuesUpdateFlag struct

struct ValuesUpdateFlags{FunDiffOrder, GeoDiffOrder, DetJdV} end
function ValuesUpdateFlags(ip_fun; update_gradients, update_hessians, update_detJdV)
    if update_gradients === update_hessians === update_detJdV === nothing
        return ValuesUpdateFlags{1, required_geo_diff_order(mapping_type(ip_fun), 1), true}() # default input till `CellValues`
    else
        FunDiffOrder = update_hessians ? 2 : (update_gradients ? 1 : 0)
        GeoDiffOrder = max(required_geo_diff_order(mapping_type(ip_fun), FunDiffOrder), update_detJdV)
        return ValuesUpdateFlags{FunDiffOrder, GeoDiffOrder, update_detJdV}()
    end
end

[codecov]

Codecov Report

Attention: Patch coverage is 97.47899% with 3 lines in your changes are missing coverage. Please review.

Project coverage is 93.69%. Comparing base (5e894e4) to head (7de2176).
Report is 2 commits behind head on master.

Files	Patch %	Lines
src/FEValues/FunctionValues.jl	94.54%	3 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #938      +/-   ##
==========================================
- Coverage   93.71%   93.69%   -0.03%     
==========================================
  Files          36       36              
  Lines        5313     5438     +125     
==========================================
+ Hits         4979     5095     +116     
- Misses        334      343       +9     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[KnutAM]

Checked the derivations - nice job! Didn't spot anything being off. Only a few suggestions for a bit clearer formatting.

[KnutAM]

Very nice @lijas.

[KnutAM]

LGTM after fixing these comments + the CI format/imports complaints 🚀

I suggest documenting this option as experimental simply to be a bit defensive on committing to the keyword interface before we have tested that a bit in practice (since it inevitabily leads to a type-unstable constructor when provided). But perhaps I'm overly careful here?

[KnutAM]

Suggested change

	CellValues([::Type{T},] quad_rule::QuadratureRule, func_interpol::Interpolation, [geom_interpol::Interpolation]; update_detJdV=true, update_gradients=true, update_hessians=false)
	CellValues([::Type{T},] quad_rule::QuadratureRule, func_interpol::Interpolation, [geom_interpol::Interpolation])

[KnutAM]

This gives a warning because it is deprecated, why is this teste removed and there is just a function call?