Feature/spatial

[cneyens]

Introducing S3 conversions from spatial objects to RMODFLOW objects (rmf_list and rmf_2d/3d/4d_arrays) and vice versa. Supported spatial classes are sf, stars and raster objects. stars is the main driver because it can handle rotated grids or rectilinear grids. Vector/raster I/O is not handled by RMODFLOW; only conversions. Most functionality is explained in the vignette 'spatial.Rmd'.

A brief summary:

• Projection information is handled through the prj object. It consists of an origin vector (bottomleft model corner coordinates), a rotation angle and a sf crs element. The prj object can be part of the dis object. Upon writing a DIS file to disk, RMODFLOW projection information is written to the header comments. When reading in a DIS file, RMODFLOW projection information is obtained from either RMODFLOW style projection information in the DIS headers, ModelMuse style projection information or OWHM style. It can also be read from a usgs.model.reference file
• By default, all functions that have a prj argument obtain this from the supplied dis argument using rmf_get_prj() which returns NULL for dis objects without a prj. prj arguments can still be supplied explicitely to function calls.
• Length units are converted on the fly between RMODFLOW (dis$lenuni) and the crs in the prj object using rmfi_prj_length_multiplier (e.g. in rmf_convert_grid_to_xyz).
• Conversion from rmf_list and rmf_2d/3d/4d_arrays to sf, stars and raster (the latter only for non-rotated regular grids).
• Conversion from sf, stars and raster to rmf_list and rmf_2d/3d/4d_arrays.
• rmf_as_tibble and other functions using coordinates have been updated
• rmf_export_* functions have been removed as RMODFLOW doesn't provide I/O for spatial objects.

Notes:

• Functions have been tested for rotated regular grids, not yet for rectilinear grids with varying cell size.
• In the current implementation, rmf_as_list.sf returns multiple cells for POINT geometries located on cell boundaries using op = sf::st_intersects (default). While this is correct behaviour from a spatial point of view, typically in the MODFLOW world, point values should be connected to a single cell. E.g. a well location should not belong the multiple cells, just one. See also the vignette.

Breaking changes:

• prj$projection is replaced by prj$crs (in line with sf).
• rmf_write_prj and rmf_read_prj are replaced by internal functions since prj is read/written from the dis file (or read from usgs.model.reference) instead of individual files.
• RMODFLOW.Rmd: Test sf chunk failed in build since rmf_as_sf has changed. Current fix: chunk not evaluated.
• rmf_install: had to skip the examples for this function in the build since they returned permission-based errors on my machine.

[rogiersbart]

Looks very complete Cas. This should also provide the required functionality for making the rmf_create_* function implementations much more straightforward I think. Will not go through lengthy testing right now, and just merge.