Three enquiries a C library has no way to answer: a directory's time, how many processors, a volume's capacity

[Sunrisepeak]

Three small ones from openkal-musl 0.10.0, in descending order of how much
trouble they cause. Each is a thing every environment beneath openkal can
answer today and the interface has no word for.

1. A directory's modification time — and this one is already being worked around

kal_fs_set_modified takes a kal_file. A directory is opened by
kal_fs_open_dir, which yields a kal_dir, and there is no kal_dir form of
kal_fs_set_modified and no form taking a name. So there is no stated route
to a directory's modification time at all, while kal_fs_info reports one
perfectly well.

A consumer met this as a lock directory it stamps to refresh the lock. Measured
directly against openkal-linux:

kal_fs_open(dir, READ|WRITE) -> 12 (kal_err_is_directory)
kal_fs_open(dir, READ)       -> 0
  kal_fs_set_modified(dir)   -> 0        <-- and the time really changes

⚠️ So openkal-musl 0.10.0 opens a directory for reading and stamps it, which
works on Linux and macOS and not on Windows (FILE_NON_DIRECTORY_FILE). That is
outside what fs.h states — the interface requires KAL_OPEN_WRITE — and it
is recorded as such rather than concealed. It is a workaround and should not
have to be one.

Asked for: a kal_dir form, or kal_fs_set_modified restated on a name.

2. How many contexts can run at once — and this one is silent

openkal.task says whether contexts run in parallel
(KAL_TASK_PROP_PARALLEL) and not how many can.

⚠️ The consequence is a wrong answer with no error attached: musl's
sysconf(_SC_NPROCESSORS_ONLN) falls back to 1, so
std::thread::hardware_concurrency() answers 1 and a program sizing a pool of
workers gets one worker and no way to know. Measured: 1 here, 32 on the host.

Every other absence in this port is loud. This one is not, which is why it is
worth a word even though nothing fails.

Asked for: an enquiry beside KAL_TASK_PROP_PARALLEL, reporting 0 for "cannot
say" so that a caller can tell that apart from 1.

3. A volume's capacity

std::filesystem::space reaches statvfs, which reports ENOSYS. There is no
operation for it. Lowest priority of the three — it is a loud absence and a
program that asks gets an error it can act on.

---

Also recorded in openkal-musl's divergence table, and not asked for here because
each is a genuine absence rather than a missing word: hard links
(kal_fs_link_create makes a node whose content is a name — a symbolic link),
named pipes, and a bidirectional stream pair (kal_process_channel is a pipe in
one direction, so socketpair cannot be built from one).