Start using log messages instead of writing output to the terminal

[jeff-cohere]

This issue is just a reminder for us to replace printf and cout statements with calls to the spdlog library that @pbosler has added to EKAT. I think Pete has the most experience with this logging package of all of us, so maybe he can clue us in on the most effective ways to use it.

[pbosler]

Option 1 is to have a haero logger that could be integrated as a component log of the host model, as shown in ekat/tests/logging/multiple_log_example.cpp, or exist on its own for standalone haero. The question is how we'd pass it around the library (perhaps as part of ModalAerosolConfig?).

Option 2 is the default spdlog logger (exposed in ekat as namespace ekat::logger::Log) which requires no instantiation of a Logger object. See ekat/tests/logging/logger_tests.cpp, section "default logger."

[jeff-cohere]

We could offer a function that allows a user to set the logger object (haero::set_logger(x)?) That would let us pick an instance from a host model or create our own if none is given by the first time a message is logged.

[pbosler]

My preference is for option 1, in the long term. It's more self-contained and sustainable. However, at this point in Haero's development, option 2 may make more sense, since option 1 will imply some (probably weak) assumptions of host model interfaces. Option 2 can be called from anywhere, if you don't mind output from every MPI rank.

If you want to filter output by mpi rank, you have to choose option 1.

[overfelt]

Sounds like option 1 is the way to go.

[jeff-cohere]

For what it's worth (I'm fiddling with cross validation tools at the moment), option 2 might be a nice way to do logging in independent test programs and tools.