Allow building the MySQL driver with MariaDB libraries.

Closes #505

Fix help_drv_pgsql.t

CI: Also build with PostgreSQL support

CI: Update help_drv_mysql.t and status badge

tests: Fix failing test due to egrep

Add GitHub actions

Add option for setting zlib/zstd compression algorithms

Use `grep -E` instead of `egrep` as `egrep` now emits a warning.

```
t/opt_report_checkpoints.t: failed
--- t/opt_report_checkpoints.t
+++ t/opt_report_checkpoints.t.err
@@ -8,6 +8,7 @@
   > fi

   $ sysbench ${SBTEST_SCRIPTDIR}/oltp_read_write.lua --db-driver=mysql --mysql-dry-run --time=3 --events=0 --report-checkpoints=1,2 run | egrep '(Checkpoint report|SQL statistics)'
+  egrep: warning: egrep is obsolescent; using grep -E
   [ 1s ] Checkpoint report:
   SQL statistics:
   [ 2s ] Checkpoint report:
```

```
$ echo | egrep 'abc'
egrep: warning: egrep is obsolescent; using grep -E
$ grep --version
grep (GNU grep) 3.8
Copyright (C) 2022 Free Software Foundation, Inc.
License GPLv3+: GNU GPL version 3 or later <https://gnu.org/licenses/gpl.html>.
This is free software: you are free to change and redistribute it.
There is NO WARRANTY, to the extent permitted by law.

Written by Mike Haertel and others; see
<https://git.sv.gnu.org/cgit/grep.git/tree/AUTHORS>.
```

Command line flag to set the sslmode for PostgreSQL (Closes: #326)

The flag name and its values match libpq's sslmode connection parameter.
The default value (prefer) will first try an SSL connection; if that
fails, it will try a non-SSL connection.

Libpq documentation: https://www.postgresql.org/docs/14/libpq-connect.html#LIBPQ-CONNECT-SSLMODE

MySQL may return multiple result sets from a call to a stored
procedure (the CALL statement), or when multiple queries are specified
in a single client request. The application is supposed to process those
multiple result sets one by one with calls to `mysql_next_result()`, or
its prepared statements counterpart, `mysql_stmt_next_result()`.

Additionally, there is the `mysql_more_results()` call which allows the
application to check if more result sets are available and works for
both regular queries and prepared statements API.

One way to handle multiple results in sysbench would be consuming all
result sets silently in the MySQL driver, but that would make it
impossible for scripts to get access to individual result sets returned
by a stored procedure.

Now sysbench exposes those MySQL client API calls to the SQL API, so it
is up to the script authors to handle multiple result sets when either
stored procedures are used in a benchmark script, or multiple queries
are passed to `sql_connection:query()`:

- sql_connection:next_result()
- sql_connection:more_results()
- sql_statement:next_results()

Here is an example how multiple results can be handled in a benchmark
script:

```lua
   local rs = con:query([[CALL p1("foo")]])
   while rs ~= nil do
      -- handle the result set
      rs = con:next_result()
   end
```

Here is a prepared statement example:

```lua
   stmt = con:prepare("CALL p1(?)")
   param = stmt:bind_create(sysbench.sql.type.CHAR, 10)
   stmt:bind_param(param)
   param:set("bar")
   rs = stmt:execute()

   while rs ~= nil do
      rs = stmt:next_result()
   end
```

Fixes GH-304.

Re-enable Ubuntu Groovy as packagecloud support has been added.

This reverts commit a4a5374ed1481df7826c130a3cd9ae3680088ca5.

Disable Fedora 33 temporarily, because it is unsupported by packagecloud
t the moment.

Disable Ubuntu Groovy temporarily, because it is unsupported by
packagecloud at the moment.

The default one (with MacOS 10.13) is too old with most Homebrew
packages (including MySQL) available in the source code form only
without prebuilt bottles.

In Debian Sid the MySQL client library now depends on libzstd, so we
have to pull that dependency as well.

SQL API failed to prepare correctly SQL statements containing floating
point arguments.

Some distributions do not alias /usr/bin/python3 to /usr/bin/python

Remove the legacy 'special' distribution, because it was unscientific
and hard to explain.

Pareto and Zipfian distributions provide more clearly defined
alternatives.

By a popular request, change the default --rand-type value from
'special' to 'uniform'.

Fixes GH-329.