...or '(no version)', if version is not present.

This also changes format; the 'K: V' format is problematic as it makes
it hard to separate from other directives.  'K=V' is now the only
accepted format.

Note that essential functions should be as fast as possible so they
probably should not inspect their arguments.

Overview of changes:

 *  Added proper obsolete declarations

Tested on Fedora, praying for it to work on Debian.

Overview of changes:

 *  Added missing dependencies

 *  Removed unnecessary dependency

 *  Codebase maintenance

Overview of changes:

 *  Override SHELLFU_INCLUDE in test runs

    Otherwise modules are actually loaded from system, if they are
    installed.

 *  SHELLFU_PATH now has preference over SHELLFU_INCLUDE

    This should have been all along; SHELLFU_PATH would not be able to
    force a particular module version in cases like testing.

 *  Added shellfu() docstring

 *  Minor co

 *  Improved and updated test header

Should have been for a long time now.

Otherwise we're not really testing "this" shellfu but the installed one.

This should have been all along; if SHELLFU_PATH is not preferred,
one cannot override it eg. in test runs.

Overview of changes:

 *  Re-packaged with better granularity

    Having multi-module packages such as 'shellfu-core' limits dependency
    control on user part.  Additionally, after splitting packages
    to one-per-module, we could move some of modules to 'sh' space,
    ie. increase their usability.

 *  Removed old 'sync' module

    Having one "wait" function does not really justify its presence in
    the main shellfu repo; if anyone needs it anyway, I can provide it
    as a separate package.

 *  Renamed binaries shellfu-doc and shellfu-get to sfpath and sfdoc

    Both were kind of hard to type.  "shellfu-get" was a bit too generic
    given that it really only prints path.  Also 'sfdoc' resembles other
    doc viewers such as 'perldoc' and 'pydoc'.

 *  Moved sfdoc to main package

    sfdoc can be useful everywhere Shellfu style is used; not just in
    Shellfu modules.

 *  Moved packages to /usr space

    Eventually, if packages should reach distributions, they should be
    there anyway.  Also it makes it easier to use Shellfu in
    environments where /usr/local/bin might be missing on PATH (eg. cron
    on debian).

 *  Improved and cleaned up lot of meta-data

    From SPEC file to debian directory to mkit.ini, there have been many
    changes in meta-data.

 *  Excluded shellfu_import test from Debian build time

    This test cannot succeed because it relies on shell features ouside
    POSIX that Debian's minimalistic dash does not implement.  The test
    will have to be rewritten; the issue is tracked here:

        https://github.com/AloisMahdal/shellfu/issues/1

 *  Added 'version' sub-command to shellfu()

    Now shellfu() function can report version of a Shellfu module.
    Versioning is not mandatory but can be useful for module unit tests
    to provide more rich reports.  (E.g. TF_HEADER in external module
    shellfu-bash-saturnin makes use of this)

 *  Refactored shellfu() a bit

 *  Made sfpath respect SHELLFU_DIR from environment

 *  Added warning in sfdoc if pod2man is missing

 *  Changed license to LGPLv2

 *  Fixed pretty.debug() bug that would hide caller name

 *  Fixed test bug with debug -v and Bash 4.4+

 *  Refactored tests a bit

 *  Added build-time tests

 *  Codebase maintenance

Add missing perl and ps dependencies and remove bash (because Essential
packages are not supposed to be there).

https://github.com/AloisMahdal/shellfu/issues/1

Bash 4.4 and later does not quote array declaration when `declare -p` is
used.

In the rare case when pod2man is not available, the usage options are
limited (one can still view pages in other modes or list functions
etc.).

Older Bash would let this get away since the output is used with form:

    read a b <<<$(_pretty__get_caller)

which joined the parts into single line, which happened to be what we
want.

Switch to packaging model where each module goes into separate package.

Such model makes much more sense (after all it's also how big platforms
such as python do it).  It allows better control over dependencies and
allows re-locating modules properly according to compatibility level
(i.e. now exit.sh can go to 'sh' despite the fact that in 'core' it had
to be with other Bash-compatible modules).

Not used anywhere these days

Mostly for consistency with the other two binaries.

'sfdoc' is easier to type and resembles 'pydoc' and 'perldoc' so it's
even easier to remember.

The only purpose of the binary is to show local Shellfu path and it's
in every single Shellfu script.  Let's save people some typing.

Since our package does not install or create anything in /usr/local,
dh_usrlocal won't try to do anything.

Decorate them and treat each argument as separate message (line)

Since __SHELLFU_DIR__ is not known at install time, its value is purely
"theoretical" and we need to be able to override it.

Overview of changes:

 *  Improved and updated README.md

    Actually, the main reason for this update is to make new README
    visible sooner, since it's been significantly improved.  Most
    important change is that the introduction has been moved to own
    document.  Also Installation section has been clarified.  That and
    some tiny fixes.

 *  Made minor visual improvement to mkusage()

 *  Added tests for mkusage() '-w' option

 *  Made SHELLFU_IMPORTED into an internal variable

 *  Added some ShellCheck exceptions

Yep, test suite is growing!