This reverts commit a6d0579.

* Drop dist packaging steps and focus instead on testing.
* Merge kafka/java matrix.
* Separate pylint step, ignore errors for now.

* handle artifact downloads; patch libs on install for kafka < 1
* export env vars
* no progressbar for wget
* Add lint: target; call pytest directly in test: recipe
* Use make targets in gh workflow; use java 21; drop java helper script
* add zstandard to requirements-dev

…#2461)

Co-authored-by: micwoj92 <45581170+micwoj92@users.noreply.github.com>

Co-authored-by: Ryar Nyah <ryarnyah@gmail.com>

Co-authored-by: HalfSweet <60973476+HalfSweet@users.noreply.github.com>

Co-authored-by: Denis Otkidach <denis.otkidach@gmail.com>

Co-authored-by: Laityned <d.a.w.markus@student.tue.nl>
Co-authored-by: misha.gavela <Arfey17.mg@gmail.com>

Co-authored-by: Dave Voutila <voutilad@gmail.com>

Co-authored-by: drewdogg <drewdogg@users.noreply.github.com>

Co-authored-by: chopatate <florian.courouge@outlook.fr>

…onn.check_version (#2477)

* conn: no connection delays between dns entries; merge blacked_out w/ connection_delay
* conn: next_ifr_request_timeout_ms() returns delay until next req timeout
* Drop poll timeout reset when no in flight requests
* Do not mark conn as sending if future immediately resolves (error)
* client poll: do not set 100ms timeout for unfinished futures
* Improve metadata refresh backoff/retry -- respect connection delays
* conn: honor reconnect backoff in connection_delay when connecting
* Log connection delay for not-ready nodes in producer sender loop
* Increase default reconnect_backoff_max_ms to 30000 (30 secs)

…ass (#2481)

…2493)

…support SaslAuthenticate (#2515)

…l_kerberos_name config (#2520)

Co-authored-by: Matt Oberle <mattoberle@users.noreply.github.com>

Co-authored-by: Emmanuel <manuco@users.noreply.github.com>

…send (#2523)

…rror details) (#2524)

…kenProvider (#2525)

…roup (#2526)