Introduce an experimental python pipeline operator

[lava]

This experimental operator is intended for situations where development velocity is more important than performance; it allows the user to run arbitrary code against the (converted-to-python) contents of a table slice.

Example:

$ tenzir 'show build | python "import math; sanitizers.root = math.sqrt(2*len(tree_hash) + len(type))"'

{
  "type": "Debug",
  "tree_hash": "0a178da187a13feefa10c1f18f7f20e9",
  "assertions": true,
  "sanitizers": {
    "address": true,
    "undefined_behavior": false,
    "root": 8.306623862918075
  }
}

Remaining Work

• Make use of the extras feature to avoid installing false dependencies for the operator venvs.
• Put the venvs under a suitable cache location.
• Synchronize venv creation to prevent write races.
• Implement a "mutable pytenzir" config option.
• Wire python errors into the operator API.
• Extend the test coverage.
• Clean up the commit history.
• Improve the docs.

[tobim]

This has been a long time coming. Thanks to everyone for their input and help. Especially @lava and @jachris.

[lava]

We decided to merge this despite the tests still failing in the macOS plugins build, because the error in question

***Exception:   0.03 sec
Errors while running CTest
dyld[8010]: Library not loaded: '/usr/local/opt/simdjson/lib/libsimdjson.18.dylib'
  Referenced from: '/Users/runner/work/tenzir/tenzir/_install/lib/libtenzir.2821.0.dylib'

is caused by the rolling upgrade of Github's macOS runners, causing the plugin builds to randomly run in a different environment than the tenzir build.

Since this PR doesn't touch the other plugins' code at all and it feels silly to keep restarting the build until we get lucky, we decided to merge the PR anyways.