Fix implicit solver compilation issues

[dennisYatunin]

This PR fixes a number of compilation issues in the implicit solver infrastructure, which have been causing inference failures and allocations in the ClimaAtmos-based unit tests (and in ClimaAtmos.jl itself).

Content

• Replace all of the recursively unrolled functions used in the MatrixFields module with generatively unrolled functions, which gets rid of the stack overflow in type inference compiler errors. Some of these functions are taken from Unrolled.jl, while others build upon the functions defined there. Move all unrolled functions into Utilities.UnrolledFunctions and add some documentation. Update the Manifest.toml files to add Unrolled.jl.
• Rewrite FieldNameDict so that there are no constant Union types being passed around, which gets rid of the allocations in the prognostic EDMF unit test. This involves merging FieldMatrix and FieldMatrixBroadcasted into a single type, and also merging FieldVectorView and FieldVectorViewBroadcasted. Add the is_lazy function to distinguish between materialized and un-materialized FieldNameDicts. Update all relevant documentation.
• Modify the definitions of norm_sqr(::FieldVector) and array_type(::FieldVector) to make them type-stable for nested FieldVectors, which fixes the remaining type instabilities in a few unit tests. Nested FieldVectors can be generated while debugging a StationaryIterativeSolve that is part of a larger FieldMatrixSolverAlgorithm.
• Change some of the FieldNameSet method definitions to avoid specializations of the form where {T}, as was suggested by @charleskawczynski. This does not appear to affect compilation, but it does make the code a bit easier to read.

Invalidations

Adding Unrolled.jl changes 4 rows in the invalidations table and brings the total number of invalidations up from 5529 to 5554. On the main branch, these 4 rows are

┌─────────────────────────────────────────────────┬───────────────┬───────────────┬─────────────────┐
│ <file name>:<line number>                       │ Function Name │ Invalidations │ Invalidations % │
│                                                 │               │               │     (xᵢ/∑x)     │
├─────────────────────────────────────────────────┼───────────────┼───────────────┼─────────────────┤
│ ColorTypes/1dGw6/src/operations.jl:24           │    isequal    │      50       │        1        │
│ InlineStrings/rlLZO/src/InlineStrings.jl:606    │     string    │      41       │        0        │
│ StaticArrays/yXGNL/src/abstractarray.jl:145     │    similar    │      35       │        0        │
│ BlockArrays/4FBcw/src/abstractblockarray.jl:221 │    getindex   │       8       │        0        │
└─────────────────────────────────────────────────┴───────────────┴───────────────┴─────────────────┘

On this branch, these 4 rows are

┌─────────────────────────────────────────────────┬───────────────┬───────────────┬─────────────────┐
│ <file name>:<line number>                       │ Function Name │ Invalidations │ Invalidations % │
│                                                 │               │               │     (xᵢ/∑x)     │
├─────────────────────────────────────────────────┼───────────────┼───────────────┼─────────────────┤
│ InlineStrings/rlLZO/src/InlineStrings.jl:606    │     string    │      77       │        1        │
│ StaticArrays/yXGNL/src/abstractarray.jl:145     │    similar    │      33       │        0        │
│ Unrolled/KtWta/src/range.jl:24                  │    getindex   │      33       │        0        │
│ ColorTypes/1dGw6/src/operations.jl:24           │    isequal    │      27       │        0        │
└─────────────────────────────────────────────────┴───────────────┴───────────────┴─────────────────┘

Since the same function names appear in both tables, the increase in invalidations is probably not related to the other changes in this PR.

[dennisYatunin]

bors try

[bors]

try

Build failed:

• buildkite/climacore-ci

[dennisYatunin]

bors try

[bors]

try

Build failed:

• buildkite/climacore-ci

[dennisYatunin]

bors try

[bors]

try

Build failed:

• buildkite/climacore-ci

[dennisYatunin]

bors try

[bors]

try

Build succeeded!

The publicly hosted instance of bors-ng is deprecated and will go away soon.

If you want to self-host your own instance, instructions are here.
For more help, visit the forum.

If you want to switch to GitHub's built-in merge queue, visit their help page.

• lib-climacore-makie
• docbuild
• test-os (ubuntu-20.04)
• lib-climacore-plots
• test-os (windows-latest)
• lib-climacore-vtk
• lib-climacore-tempestremap
• buildkite/climacore-ci
• format

[dennisYatunin]

bors try

[bors]

try

Build succeeded!

The publicly hosted instance of bors-ng is deprecated and will go away soon.

If you want to self-host your own instance, instructions are here.
For more help, visit the forum.

If you want to switch to GitHub's built-in merge queue, visit their help page.

• docbuild
• test-os (ubuntu-20.04)
• test-os (windows-latest)
• lib-climacore-vtk
• lib-climacore-makie
• lib-climacore-plots
• lib-climacore-tempestremap
• buildkite/climacore-ci
• format

[dennisYatunin]

bors try

[bors]

try

Build succeeded!

The publicly hosted instance of bors-ng is deprecated and will go away soon.

If you want to self-host your own instance, instructions are here.
For more help, visit the forum.

If you want to switch to GitHub's built-in merge queue, visit their help page.

• lib-climacore-plots
• lib-climacore-makie
• test-os (ubuntu-20.04)
• test-os (windows-latest)
• lib-climacore-vtk
• lib-climacore-tempestremap
• buildkite/climacore-ci
• format
• docbuild

[dennisYatunin]

bors try

[bors]

try

Build failed:

• buildkite/climacore-ci

[dennisYatunin]

bors try

[bors]

try

Build failed:

• buildkite/climacore-ci

[charleskawczynski]

We may want to revisit this at some point to see if we can eliminate the unroll functions, but overall looks fine to me.

[charleskawczynski]

Also, we can just ignore the invalidations issue (we can revisit this when we upgrade to julia 1.10)