tests/test-arbitraryfilectx.t

phases: invalidate the phases set less often on retract boundary

We already have the information to update the phase set, so we do so directly
instead of invalidating the cache.

This show a sizeable speedup in our `perf::unbundle` benchmark on the
many-draft mozilla-try repository.

### data-env-vars.name            = mozilla-try-2023-03-22-zstd-sparse-revlog
  # benchmark.name                = hg.perf.perf-unbundle
  # bin-env-vars.hg.flavor        = no-rust
  # bin-env-vars.hg.py-re2-module = default
  # benchmark.variants.issue6528  = disabled

  # benchmark.variants.revs = last-10
before: 2.055259 seconds
after:  1.887064 seconds (-8.18%)
  # benchmark.variants.revs = last-100
before: 2.409239 seconds
after:  2.222429 seconds (-7.75%)
  # benchmark.variants.revs = last-1000
before: 3.945648 seconds
after:  3.762480 seconds (-4.64%)

Setup:
  $ cat > eval.py <<EOF
  > import filecmp
  > from mercurial import commands, context, pycompat, registrar
  > cmdtable = {}
  > command = registrar.command(cmdtable)
  > @command(b'eval', [], b'hg eval CMD')
  > def eval_(ui, repo, *cmds, **opts):
  >     cmd = b" ".join(cmds)
  >     res = pycompat.bytestr(eval(cmd, globals(), locals()))
  >     ui.warn(b"%s" % res)
  > EOF

  $ echo "[extensions]" >> $HGRCPATH
  $ echo "eval=`pwd`/eval.py" >> $HGRCPATH

Arbitraryfilectx.cmp does not follow symlinks:
  $ mkdir case1
  $ cd case1
  $ hg init
#if symlink
  $ printf "A" > real_A
  $ printf "foo" > A
  $ printf "foo" > B
  $ ln -s A sym_A
  $ hg add .
  adding A
  adding B
  adding real_A
  adding sym_A
  $ hg commit -m "base"
#else
  $ hg import -q --bypass - <<EOF
  > # HG changeset patch
  > # User test
  > # Date 0 0
  > base
  > 
  > diff --git a/A b/A
  > new file mode 100644
  > --- /dev/null
  > +++ b/A
  > @@ -0,0 +1,1 @@
  > +foo
  > \ No newline at end of file
  > diff --git a/B b/B
  > new file mode 100644
  > --- /dev/null
  > +++ b/B
  > @@ -0,0 +1,1 @@
  > +foo
  > \ No newline at end of file
  > diff --git a/real_A b/real_A
  > new file mode 100644
  > --- /dev/null
  > +++ b/real_A
  > @@ -0,0 +1,1 @@
  > +A
  > \ No newline at end of file
  > diff --git a/sym_A b/sym_A
  > new file mode 120000
  > --- /dev/null
  > +++ b/sym_A
  > @@ -0,0 +1,1 @@
  > +A
  > \ No newline at end of file
  > EOF
  $ hg up -q
#endif

These files are different and should return True (different):
(Note that filecmp.cmp's return semantics are inverted from ours, so we invert
for simplicity):
  $ hg eval "context.arbitraryfilectx(b'A', repo).cmp(repo[None][b'real_A'])"
  True (no-eol)
  $ hg eval "not filecmp.cmp(b'A', b'real_A')"
  True (no-eol)

These files are identical and should return False (same):
  $ hg eval "context.arbitraryfilectx(b'A', repo).cmp(repo[None][b'A'])"
  False (no-eol)
  $ hg eval "context.arbitraryfilectx(b'A', repo).cmp(repo[None][b'B'])"
  False (no-eol)
  $ hg eval "not filecmp.cmp(b'A', b'B')"
  False (no-eol)

This comparison should also return False, since A and sym_A are substantially
the same in the eyes of ``filectx.cmp``, which looks at data only.
  $ hg eval "context.arbitraryfilectx(b'real_A', repo).cmp(repo[None][b'sym_A'])"
  False (no-eol)

A naive use of filecmp on those two would wrongly return True, since it follows
the symlink to "A", which has different contents.
#if symlink
  $ hg eval "not filecmp.cmp(b'real_A', b'sym_A')"
  True (no-eol)
#else
  $ hg eval "not filecmp.cmp(b'real_A', b'sym_A')"
  False (no-eol)
#endif