Doc: Add MSSQL Driver examples

[geographika]

Add some MSSQL driver examples, including PowerShell and new CLI. Tested on Windows and Linux.

One point of confusion was if I specified the --layer in the read step of a pipeline I got errors reporting the schema name:

# ERROR 1: Unable to open secondary datasource geo required by JOIN

Example:

  gdal vector pipeline \
      ! read "$conn" --layer "geo.rivers" \
      ! sql --sql "SELECT oid, geom FROM geo.rivers WHERE hydroarea=10" \
      ! write out.gpkg --overwrite

Removing --layer resolved this, so I added it to the docs.

[rouault]

One point of confusion was if I specified the --layer in the read step of a pipeline I got errors reporting the schema name:

trying a bit the same with a shapefile or postgis source, I actually got crashes... Now fixed by bcac881 . But your issue was a bit different. --layer caused ExecuteSQL() to run with OGRSQL instead f MSSQL. Should be fixed per 59e4e1f

[geographika]

Thanks @rouault. I'll convert this PR to a draft, retest, and then remove that line from the example.

I did run into another cryptic error (that seemed intermittent):

ERROR 1: An error occurred while writing a dirty block from DropCache

Any ideas what this could mean (and if it should be documented?)

The issue was the account used in $conn2 didn't have trusted access, and I did also get ERROR 1: Unable to initialize connection to the server for ....

gdal vector pipeline \
  ! read $conn1--layer "input.test" \
  ! select --fields "Geom" --output-layer "lyr" \
  ! limit --limit 100 \
  ! sql --sql "SELECT 1 Id, Geom FROM lyr" \
  ! write $conn2 --append --output-layer "output.test"

I also have a note that I got this message when I had the wrong name for a filter step, but I couldn't recreate this.

[rouault]

Any ideas what this could mean (and if it should be documented?)

this is likely a bug. This is related to the FlushCache() raster logic... Definitely not something to document, but to fix. Any way to reproduce with other datasources than MSSQL... ?

[geographika]

Any ideas what this could mean (and if it should be documented?)

this is likely a bug. This is related to the FlushCache() raster logic... Definitely not something to document, but to fix. Any way to reproduce with other datasources than MSSQL... ?

I'll try and recreate with file datasets (it will be at the end of the month). There was definitely no raster data involved though.