Add support for spdlog

[eddelbuettel]

This (draft) PR adds support for spdlog. On Windows, it is currently disabled as the linking with libspdlog.a (to support libtiledb) creates an issue.

Edit: This has been reworked, and the logging ability is now exported directly from the RcppSpdlog package. It requires a version that is slightly ahead of what CRAN has; CI is installing this explicitly right now.

Having this support allows to control logging both from the package and (via the standard config object) at the TileDB Core level:

$ r -ltiledb,RcppSpdlog -e'log_set_level("info"); set_allocation_size_preference(1024^3); \
      cfg <- tiledb_config(); \
      cfg["config.logging_level"] <- 4; \
      ctx <- tiledb_ctx(cfg); \
      print(tiledb_array("../tiledb-soma/test/soco/pbmc3k_processed/ms/mRNA/X/data", query_layout="UNORDERED")[])'
[2022-10-28 17:31:46.814] [tiledb] [Process: 1101416] [info] [tiledb_array] query: READ
[2022-10-28 17:31:46.820] [tiledb] [Process: 1101416] [info] [tiledb_array] '[' accessor started
[2022-10-28 17:31:46.945] [tiledb] [Process: 1101416] [info] [libtiledb_query_submit]
[2022-10-28 17:31:46.948] [Process: 1101416] [debug] [1666996306731388389-Context: 1] [Query: 1] [SparseGlobalOrderReader: 1] Initial data loaded
[2022-10-28 17:31:46.949] [Process: 1101416] [debug] [1666996306731388389-Context: 1] [Query: 1] [SparseGlobalOrderReader: 1] Done adding result tiles, num result tiles 49
[2022-10-28 17:31:46.949] [Process: 1101416] [debug] [1666996306731388389-Context: 1] [Query: 1] [SparseGlobalOrderReader: 1] All result tiles loaded
[2022-10-28 17:31:46.983] [Process: 1101416] [debug] [1666996306731388389-Context: 1] [Query: 1] [SparseGlobalOrderReader: 1] Done reading and unfiltering coords tiles
[2022-10-28 17:31:46.992] [Process: 1101416] [debug] [1666996306731388389-Context: 1] [Query: 1] [SparseGlobalOrderReader: 1] Done computing tile bitmaps
[2022-10-28 17:31:46.992] [Process: 1101416] [debug] [1666996306731388389-Context: 1] [Query: 1] [SparseGlobalOrderReader: 1] Done applying query condition
[2022-10-28 17:31:46.993] [Process: 1101416] [debug] [1666996306731388389-Context: 1] [Query: 1] [SparseGlobalOrderReader: 1] Done processing fragments with timestamps
[2022-10-28 17:31:47.004] [Process: 1101416] [debug] [1666996306731388389-Context: 1] [Query: 1] [SparseGlobalOrderReader: 1] Done merging result cell slabs, num slabs 49, buffers full false
[2022-10-28 17:31:47.049] [Process: 1101416] [debug] [1666996306731388389-Context: 1] [Query: 1] [SparseGlobalOrderReader: 1] Done copying tiles, buffers full false
[2022-10-28 17:31:47.050] [Process: 1101416] [debug] [1666996306731388389-Context: 1] [Query: 1] [SparseGlobalOrderReader: 1] Done with iteration, num result tiles 0
[2022-10-28 17:31:47.050] [tiledb] [Process: 1101416] [info] [libtiledb_query_status]
[2022-10-28 17:31:47.050] [tiledb] [Process: 1101416] [info] [tiledb_array] '[' query status:COMPLETE
         soma_dim_0 soma_dim_1  soma_data
      1:          0          0 -0.1714695
      2:          0          1 -0.2808120
      3:          0          2 -0.0466768
      4:          0          3 -0.4751688
      5:          0          4 -0.5440238
     ---                                 
4848640:       2637       1833 -0.1038306
4848641:       2637       1834 -0.4368798
4848642:       2637       1835 -0.0784244
4848643:       2637       1836 -0.1303272
4848644:       2637       1837 -0.4713379
$ 

The question to discuss now is whether a) this useful as-is on the two main OSs we use for development and b) whether it may be worth rejigging the core library build to export the logger (which would clearly be a more medium-term change).

[shortcut-integration]

This pull request has been linked to Shortcut Story #22999: Integrate spdlog.

[eddelbuettel]

Did one more round of tests and am now quite positive this works as intended -- giving us spdlog on all platforms without any double symbol woes. It currently requires RcppSpdlog 0.0.8.4 as in the ghrr drat but I will ship that to CRAN 'soon'.