Maintenance swipe: get rid of some global variables in favour of actual data flow visibility

[jnpkrn]

This is a first step in a direction towards less rigid/cumbersome
code, easing the understandability and maintenance as such.

Get rid of (some, commonly shared) global variables means:

• less hidden, behind the curtains state, meaning less
  of surprising side-effects, better understandable data
  flow, effectively containing the impact radius of the
  functions

• unit tests are more feasible (easier to conduct) when the
  inputs to the computation/functions are not part of said
  non-apparent global state, but rather part of the explicit
  input to the them

• likewise, any "scope of authority" etc. types of regrouping
  of the functions in modules is not easily done when the
  data flow is partially obfuscated

• more dynamic handling is enabled (e.g. multithreading or
  fetching a new instance of the configuration, which can then
  be conditionally promoted to the life configuration)
  if there's a good need for that

• last but not least, using global-scoped variables is not
  considered a praised practice

For struct booth_config in particular, it was dynamically
alloc'd at heap, so there was no gain in terms of avoiding
relevant class of bugs.

Also, this PR reconciles the no-code-doc problem at the functions
being touched throughout. And this swipe helped to identify
both unused and needlessly exported functions.

[jnpkrn]

And not to forget ... current passing of the whole
struct booth_config is an apparent overapproximation,
where just some bits of that structure are needed inputs
for some particular functions to operate upon.

I haven't delved into that business yet, but it's one
of the most immediate steps to follow. Note that without
this rather radical intervention, such a fine-graining
would hardly be possible.

[jnpkrn]

Wrt. fine-graining, desired access method (read-only/ok-to-modify) for
subset of that struct can now finally be expressed/guarded -- something
you can't viably achieve with an arbitrary-touch-for-everything global.

[jnpkrn]

So it imposes structural (vs. hazy flat) patterns on multiple levels,
which is desirable.

[jnpkrn]

Some previous commits amended, some new added.

Known issue:

boothd: transport.c:153: _find_myself: Assertion `conf_ptr->local != NULL' failed.

This is because of intentionally crafted asserts to discover exactly
this sort of naively expected guarantees.

Need to look into the call chain whether it's not malign, to begin with.

[dmuhamedagic]

It is indeed less than optimal for the booth_config to be shared as is. Thanks for bringing this up.
It would, however, be great if could have a better separation of concerns. I guess that we would automatically get a more localized data access. For instance, the consensus algorithm is run on a per ticket basis, which is most probably an overkill and it should be run on a site basis. That's the most glaring issue, but there could be more.
Splitting booth_config is very much worth considering, as you already mentioned.

[dmuhamedagic]

Hi Jan, apparently you are not involved any more with HA. Wanted to say that it is a pity that we don't have your help anymore, but I guess that it was not easy fighting against proverbial wind mills. It is my fault that I often did not have time to work on your pull request, though heaven knows that I really tried. Occasionally, it would be just too much for me. But that doesn't mean that your PRs were bad, to the contrary, they were always of excellent quality and your engagement was always very much appreciated.

[dmuhamedagic]

Having said all that, I hope that you are doing fine, wherever you work now. Cheers, Dejan (-:

[clumens]

Don't mind me, just making a checklist so I know which patches I've merged and which I still have left to do.

• no fear from a word interfere
• foreach_{node,ticket}: apply more universally
• rename foreach_{node,ticket} macros to uppercased variants
• read_config to no longer rely on global booth_conf variable
• check_config to no longer rely on global booth_conf variable
• find_site_by_name to no longer rely on global booth_conf var
• find_site_by_id to no longer rely on global booth_conf var
• find_myself to no longer rely on global booth_conf variable
• setup_udp_server to no longer rely on global booth_conf var
• FOREACH_{NODE,TICKET} to no longer rely on global booth_conf
• find_ticket_by_name to no longer rely on global booth_conf var
• rest of ticket.c to no longer rely on global booth_conf var
• notify_client to no longer rely on global booth_conf var
• rest of transport.c to no longer rely on global booth_conf var
• rest of raft.c to no longer rely on global booth_conf var
• rest of inline-fn.h to no longer rely on global booth_conf var
• get rid of any use of booth_conf global var except for main.c
• get rid of any sparable use of booth_conf global var in main.c
• get rid of booth_transport as a global var&finish booth_config
• get rid of pcmk_handler as a global variable
• get rid of "struct command_line" as a global variable
• get rid of poll_timeout global variable
• get rid of loval and no_leader and global variables
• some more cleanups (unused or shall-be-static functions)
• factor safe_copy function to an auxiliary utils module
• move check_max_len_valid function to existing utils module
• move find_ticket_by_name to config module, for a better git
• systemic disposal of struct booth config