chore: migrate to conan 2 (#440)

* chore: bump to conan2

* gha: fix conan-cache-action

* chore: run python formatting

* chore: more formatting fixes

* chore: fix compilation warnings

* gha: fix conan caching

* chore: bump protobuf version

* ci: run dev.conan to generate the generators

* fix(dist-test): build conan first

* fix(gha): temporarily disable TSAN and ASAN

* fix: use different conan profiles for sanitisers

* fix: run cpp formatting

* fix(conan): update conan-release.lock

* fix(ci): add conan toolchain to example's build

* fix(gha): don't force --clean on the conan-cache step

chore: bump images to ubuntu 24.04 (#438)

* git: bump code version to 0.21.0

* util: add cstdint header

* docker: bump planner base version to right faabric-base

threads: elastically scale-up (#418)

* threads: elastically scale-up

* gh: bump code version

* threads: handle zero-sized requests and preload a la mpi

* omp: fix race conditions with elastic scaling

planner: introduce SPOT VMs policy (#433)

planner: implement compaction planner policy (#430)

planner(mpi): preload MPI SCALE_CHANGE decisions (#422)

* planner(mpi): preload MPI SCALE_CHANGE decisions

When scheduling a NEW app with the `ismpi` flag set, we know that this
request will be quickly followed-up by a SCALE_CHANGE request with
`worldSize` - 1 messages. As a consequence, and to guarantee optimal
bin-packing on-par with Slurm, we can schedule `worldSize` messages when
the first request arrives, dispatch just the first one, and preload the
`worldSize` - 1 remaining.

* deps(python): bump sphinx version

gh: bump minor version to 0.16.0 (#404)

LLVM 17 (#376)

* gh: bump code version

* docker: bump to llvm 17

* tasks(dev): read clang version from env var

* cmake: conan fixes

* tasks: fix llvm major version usage

* fix: fixes to build faasm with llvm 17

* docker(planner): update planner's base image

* docker: bump to llvm 18

* conan: fixes to use llvm 18

* llvm: try using version 17 for c++ compilation

* gha: temporarily skip formatting checks

* gha: remove warnings in release build

* tasks: pin llvm version major to variable

* clang-format: fix braces formatting after llvm bump

* nits: run clang-format-17

* nits: self-review

* cmake: remove unnecessary comment

* scheduler: reset when re-using a warm scheduler

* nits: self-review

planner(migration): support migrating to a completely different node (#…

…369)

* planner(migration): support migrating to a completely different node if that reduces cross-vm links

* chore: bump code version

batch-scheduler: fix the number of cross-vm links computation (#367)

* batch-scheduler: fix the number of cross-vm links computation

* chore: bump patch code version