- this bumps format_version of tinshift JSON to 1.1 for the new field
   fallback_strategy
 - the default behaviour without that field is retained
 - if fallback_strategy is set to "nearest_side", then points that do not fall
   into any of the triangles will be transformed according to the nearest
   triangle
 - if fallback_centroid is set to "nearest_side", then points that do not fall
   into any of the triangles will be transformed according to the triangle
   with the nearest centroid

This makes it easier to turn off all programs, rather than individually.
Useful for example to avoid https://github.com/OSGeo/gdal/blob/master/gdal/fuzzers/build.sh#L138

Fixes #2779

proj_factors(): accept P to be a projected CRS (fixes #2854)

Updated doc:

    Starting with PROJ 8.2, the P object can be a projected CRS, for example
    instantiated from a EPSG CRS code. The factors computed will be those of the
    map projection implied by the transformation from the base geographic CRS of
    the projected CRS to the projected CRS.

    The input geodetic coordinate lp should be such that lp.lam is the longitude
    in radian, and lp.phi the latitude in radian (thus independently of the
    definition of the base CRS, if P is a projected CRS).

DOC: Add content for Development/Error handling page.

* DOC: Refresh Development/Quickstart and remove Development/Threads

The document Development/Threads was severely out of date, and was
focussed on contrasting the current API with an old API that was
deprecated several years ago, and removed from the current PROJ earlier
this year.

Update Development/Quickstart to give more information about the use of
threading context, as a replacement for the content that was previously
in Threads.

Also give the example code examples/pj_obs_api_mini_demo.c a cleanup
pass to make it at least internally consistent with its own code style.

The header comments of the example code are still, much like the Threads
document, fixated on comparing proj.h against proj_api.h, which is an
interesting historical curio to be sure, but probably doesn't need to
persist here.  It might be worth cleaning this up further as a separate
exercise.

Fixes #2451