[README] Fibirl uses spawn/sync as keywords (and no longer fork/join)

See merge request i4/manycore/emper!403

[io/Stats] For AlarmFutures (TIMEOUT) expect them to return ETIME

See merge request i4/manycore/emper!402

[Makefile] Add asan(-test) targets

See merge request i4/manycore/emper!398

LTO

See merge request i4/manycore/emper!401

This prevents a linker error because of undefined symbols when LTO is
used.

[Runtime] Switch logging of env var handling from DBG() to LOGI()

See merge request i4/manycore/emper!400

Also use std::boolalpha to print ture/false.

[SpuriousFutex2Semaphore] Add missing emper-config.h include

See merge request i4/manycore/emper!399

Fixes: f4e72a97 ("[iwyu] Fix includes as per iwyu 0.18")

[ContextManager] Remove the put/get free context cycle

See merge request i4/manycore/emper!396

The put/get free context cycle in ContextManager::start() (if
continuation stealing is enabled) became unnecessary with
c66551bf ("[Context] Use 'jmp' instead of 'ret' to kickoff the
context").

Fix ContextManager with MemoryManager

See merge request i4/manycore/emper!395

Even though the MemoryManager is enabled for the ContextManager per
default, it never was actually used due a logic bug. Squash that bug!

In case a Context has been retrieved from the MemoryManager, without
it malloced, then we need to re-initialze some of the Context's values
before we can use the Context.

Kickoff context using 'jmp' instead of 'ret'

See merge request i4/manycore/emper!394

My system running 5.15.43 also returns ECANCELED.

[SemaphoreWorkerSleepStrategy] Improve code comment

See merge request i4/manycore/emper!385

[gitlab-ci] Bump debian-testing-dev image to 1.23

See merge request i4/manycore/emper!391

Newer iwyu versions will exit with an error code in case include
issues have been found.

echoclient: initialize variable to fix building with gcc 12.1

See merge request i4/manycore/emper!388

pulse: only measure time between work creation and dispatching it

See merge request i4/manycore/emper!386

runtime: do not notify workers about termination if they do not sleep

See merge request i4/manycore/emper!387

I stumbled upon this because the no-sleep variants in the pulse evaluation
showed notifications in their stats.

We do not need to measure the work duration which is the known pulse.
This makes comparing the raw data easier because the latency does not
include the pulse duration.

fsearch: add possible semaphore limiting the concurrenct open requests

See merge request i4/manycore/emper!382