Adopt Python 3.10 in the Haystack ecosystem

[anakin87]

As part of the move to Python 3.10, we plan to:

• testing with 3.10 on CI
• dropping 3.9 from pyproject
• adopting X|Y typing syntax (PEP604) while maintaining compatibility with previous syntax

What this means for the ecosystem 👇

Haystack (#9854)

• testing with 3.10 on CI (chore!: drop Python 3.9 support due to EOL #10244 ✅)
• dropping 3.9 from pyproject (chore!: drop Python 3.9 support due to EOL #10244 ✅, to be released in 2.22.0)
• adopting X|Y typing syntax (PEP604): (refactor: support PEP604 typing syntax (X|Y and X|None)  #10255 ✅, to be released in 2.22.0)

Haystack Core integrations

• testing with 3.10 on CI (test: use Python 3.10 in tests instead of 3.9 haystack-core-integrations#2616 and test: use Python 3.10 in tests instead of 3.9 - pt. 2 haystack-core-integrations#2620 ✅)
• dropping 3.9 from pyproject: TODO after 2.22.0 release
• adopting X|Y typing syntax: TODO after 2.22.0 release

Notes:
The last two tasks can be done together. Since the X|Y typing syntax will only be supported starting from 2.22.0, this implies for each integration we should:

• open an update PR
• merge it
• release a new major version that pins haystack-ai>=2.22.0

Haystack Experimental

• test with 3.10: (chore: test with python 3.10 in the CI haystack-experimental#415 ✅)
• dropping 3.9 from pyproject: TODO after 2.22.0 release
• adopting X|Y typing syntax: TODO after 2.22.0 release

Notes:

• After completing the first step, it is possible that tests against Haystack main may fail. This could happen because some components overridden in Experimental might rely on the new typing syntax introduced in Haystack. If this happens, I would keep these failures until 2.22.0 release.
• The second and the third step can be done together. I propose not pinning a specific version of Haystack in Experimental. While more precise, this could worsen cyclic dependency (Haystack depends on Experimental, which depends on Haystack), leading to fragile/confusing dependency resolution, as we have seen in the past.

[anakin87]

@sjrl I'd appreciate it if you could take a look and confirm the planned approach, especially about Core Integrations and Experimental.

[sjrl]

@anakin87 thanks for putting this plan together and the plan for both core-integrations and experimental look good to me!