tests/test-revisions.t

make: switch the PYTHON default to `py.exe -3` on Windows

Python3 _is_ named `python.exe` on Windows, but that isn't necessarily on PATH
when installing from python.org.  I do happen to have a python.exe on PATH in
`$LOCALAPPDATA/Microsoft/WindowsApps`, but it appears to be 0 bytes (likely
because of permission issues), and doesn't run:

    $ python -V
     - Cannot open

Pulkit hit the same error as I did though, so it isn't just my system:

    $ make -C . local
    make: Entering directory `/home/Dell/repos/hg-committed`
    python setup.py  \
              build_py -c -d . \
              build_ext  -i \
              build_hgexe  -i \
              build_mo
     - Cannot openmake: *** [local] Error 1

The `py.exe` dispatcher lives in the Windows directory (so it is on PATH), looks
up the python.org installation, and invokes that interpreter directly.  I get a
warning with py39, but if it's our issue, it was an existing one:

    $ make -C .. local
    make: Entering directory `/c/Users/Matt/hg'
    py -3 setup.py  \
              build_py -c -d . \
              build_ext  -i \
              build_hgexe  -i \
              build_mo
    C:\Users\Matt\AppData\Local\Programs\Python\Python39\lib\site-packages\setuptools\distutils_patch.py:25:
        UserWarning: Distutils was imported before Setuptools. This usage is discouraged and may
        exhibit undesirable behaviors or errors. Please use Setuptools' objects directly or at least
        import Setuptools first.
      warnings.warn(

The end result is a py3 based hg.exe that annoyingly won't run because it can't
find python39.dll.  It will run tests (the ones without the `python3` shbang
line anyway), because the test runner adjusts PATH to include the python running
it.

Differential Revision: https://phab.mercurial-scm.org/D9361

  $ hg init repo
  $ cd repo

  $ echo 0 > a
  $ hg ci -qAm 0
  $ for i in 5 8 14 43 167; do
  >   hg up -q 0
  >   echo $i > a
  >   hg ci -qm $i
  > done
  $ cat <<EOF >> .hg/hgrc
  > [alias]
  > l = log -T '{rev}:{shortest(node,1)}\n'
  > EOF

  $ hg l
  5:00f
  4:7ba5d
  3:7ba57
  2:72
  1:9
  0:b
  $ cat <<EOF >> .hg/hgrc
  > [experimental]
  > revisions.disambiguatewithin=not 4
  > EOF
  $ hg l
  5:00
  4:7ba5d
  3:7b
  2:72
  1:9
  0:b
9 was unambiguous and still is
  $ hg l -r 9
  1:9
7 was ambiguous and still is
  $ hg l -r 7
  abort: 00changelog.i@7: ambiguous identifier!
  [255]
7b is no longer ambiguous
  $ hg l -r 7b
  3:7b

  $ cd ..