fix(session): create honours an absolute path whose parent exists

[russimicro]

Follow-up to #142, from the same deployment.

Problem

#142 honoured an absolute create path when it sits under data_dir_. The
other half of the same report is an application that stages work tables
somewhere else and reads them back by the same absolute name — ours builds
scratch copies under the TEMP drive.

Those paths still fold:

} else if (path_is_inside(data_dir_, rel)) {
    return platform::resolve_case_insensitive(rel.string());
}
rel = rel.relative_path();          // <- everything else lands here

and the fold produces data_dir_ / <the whole client path>, whose intermediate
directories nobody creates. The create fails, and the caller cannot find its
table under either name.

Change

Also honour an absolute path when its parent directory already exists. That
is what the real ACE engine does for a free-table create: it writes exactly the
file it was given.

Deliberately unchanged:

• a path whose parent does not exist still folds — that is the case the
  guard was written for, a client passing a path from its own working directory
  that means nothing on the server;
• a bare drive-root name (C:\STRAY.DBF) still folds — the root always exists,
  but writing a table there is never a deliberate location, it is the signature
  of a client that built the name from its own root. abi_create_table_test
  pins that and still passes.

Verification

tests/unit/abi_create_outside_datadir_test.cpp — creates into an existing
directory outside data_dir_, checks the file lands there and not under
data_dir_, and reopens it by the same absolute name. Second subcase checks a
non-existent parent still does not land at the caller's path.

Full suite on this branch (MSVC x64 Release): 1228/1235 — the 7 failures are
pre-existing on main and unrelated (6 × abi_pritpal_lock_test remote
connect, 1 × abi_server_fs_test teardown that threw on remove_all because
the temp directory was held by another process on this machine).

[gemini-code-assist]

Caution

The consumer version of Gemini Code Assist on GitHub has been sunset. All code review activity has officially ceased.

[FiveTechSoft]

Integrated into main as part of the PR stack merge (9e7ac41). Cherry-picked onto post-CI-fix main, conflict-resolved (CHANGELOG / CMakeLists), and verified locally. Thank you @russimicro.