NaN and Inf issues in exp

The test in https://github.com/JuliaArrays/StaticArrays.jl/blob/59f92e0ca7ac391a850a6e7a2ce1eb53aa237fc4/src/expm.jl#L21-L32 don't cover the NaN case so

julia> A = @SMatrix([1.0 0.1; NaN 1.0])
2×2 SArray{Tuple{2,2},Float64,2,4} with indices SOneTo(2)×SOneTo(2):
   1.0  0.1
 NaN    1.0

julia> exp(A)
2×2 SArray{Tuple{2,2},Float64,2,4} with indices SOneTo(2)×SOneTo(2):
   2.71828  0.271828
 NaN        2.71828

which should either throw or give a NaN. Also, https://github.com/JuliaArrays/StaticArrays.jl/blob/59f92e0ca7ac391a850a6e7a2ce1eb53aa237fc4/src/expm.jl#L96-L99 can be reached for matrices with Infs which will error in an uninformative way

julia> A = @SMatrix([1.0 0.1 0.0; Inf 1.0 0.1; 0.1 0.0 1.0])
3×3 SArray{Tuple{3,3},Float64,2,9} with indices SOneTo(3)×SOneTo(3):
  1.0  0.1  0.0
 Inf   1.0  0.1
  0.1  0.0  1.0

julia> exp(A)
ERROR: InexactError: trunc(Int64, Inf)
Stacktrace:
 [1] trunc at ./float.jl:703 [inlined]
 [2] ceil at ./float.jl:365 [inlined]
 [3] _exp(::Size{(3, 3)}, ::SArray{Tuple{3,3},Float64,2,9}) at /Users/vagrant/.julia/packages/StaticArrays/LJQEe/src/expm.jl:97
 [4] exp(::SArray{Tuple{3,3},Float64,2,9}) at /Users/vagrant/.julia/packages/StaticArrays/LJQEe/src/expm.jl:1
 [5] top-level scope at REPL[70]:1
 [6] eval(::Module, ::Any) at /Applications/Pumas-1.1.app/Contents/Resources/julia/Contents/Resources/julia/lib/julia/sys.dylib:?
 [7] eval_user_input(::Any, ::REPL.REPLBackend) at /Users/vagrant/worker/juliapro-staging-osx1011-0_6/build/tmp_julia/Julia-1.5.app/Contents/Resources/julia/share/julia/stdlib/v1.5/REPL/src/REPL.jl:134
 [8] repl_backend_loop(::REPL.REPLBackend) at /Users/vagrant/worker/juliapro-staging-osx1011-0_6/build/tmp_julia/Julia-1.5.app/Contents/Resources/julia/share/julia/stdlib/v1.5/REPL/src/REPL.jl:195
 [9] start_repl_backend(::REPL.REPLBackend, ::Any) at /Users/vagrant/worker/juliapro-staging-osx1011-0_6/build/tmp_julia/Julia-1.5.app/Contents/Resources/julia/share/julia/stdlib/v1.5/REPL/src/REPL.jl:180
 [10] run_repl(::REPL.AbstractREPL, ::Any; backend_on_current_task::Bool) at /Users/vagrant/worker/juliapro-staging-osx1011-0_6/build/tmp_julia/Julia-1.5.app/Contents/Resources/julia/share/julia/stdlib/v1.5/REPL/src/REPL.jl:292
 [11] run_repl(::REPL.AbstractREPL, ::Any) at /Users/vagrant/worker/juliapro-staging-osx1011-0_6/build/tmp_julia/Julia-1.5.app/Contents/Resources/julia/share/julia/stdlib/v1.5/REPL/src/REPL.jl:288
 [12] (::Base.var"#807#809"{Bool,Bool,Bool,Bool})(::Module) at ./client.jl:399
 [13] run_main_repl(::Bool, ::Bool, ::Bool, ::Bool, ::Bool) at /Applications/Pumas-1.1.app/Contents/Resources/julia/Contents/Resources/julia/lib/julia/sys.dylib:?
 [14] exec_options(::Base.JLOptions) at /Applications/Pumas-1.1.app/Contents/Resources/julia/Contents/Resources/julia/lib/julia/sys.dylib:?
 [15] _start() at /Applications/Pumas-1.1.app/Contents/Resources/julia/Contents/Resources/julia/lib/julia/sys.dylib:?

Start in src/expm.jl, especially the existing test area at lines 21–32 and the Inf-related code at lines 96–99. Run the NaN and Inf matrix examples from the issue, then inspect how exp handles these values. Done means the behavior is covered by tests and no longer produces the reported uninformative result.