Don't worry about it.
Like -- at all!
Place the components, surround with ground, drop a few stitching vias, keep traces a modest distance away from them, and you're done.
Your first layout is perfectly fine, as far as what is shown. If you want to avoid crossing the trace with one capacitor, rotating the one is fine. If this is a commercially fabbed board, traces under components will not violate any clearance issues, and soldermask will be effective insulation for this voltage. Ground keepout is unnecessary: you're literally putting extra capacitors to ground around the crystal, what difference would another ~0.8pF or so make to it? But neither will it hurt, so my above statement remains true.
---
I'm always a bit amused by crystal layout, because it's such a common, and frequent, topic for bikeshedding.
Bikeshedding: definitionally--see the responses in this thread. Do this, do that, move components here, put traces there. It's
entirely superficial:: what the bike shed should look like, put this siding here, put that window there, paint it with this or that color.
No one is discussing, or has yet even mentioned, any of the hard questions, like: "Where do we put the bike shed?" "How big should it be?" "How much will it cost and how will we budget for it?" "What permits if any do we need to build it?" That is: any matters of material importance, functionally necessary to actually build a bike shed.
Hence the origin and meaning of this term. I don't think there was a literal, formative instance of a proverbial bike shed, but the example rings true time and again, as individuals or groups debate the easy and superficial aspects of a problem and leave the hard and awkward questions unturned.
Nothing that has been mentioned will have a measurable impact on EMC performance, or more than a dB's worth. It's just not that important of a thing to spend so much time or words on, and more importantly: anything that would make a difference, is an actual
hard problem -- you need detailed datasheets for both MCU and crystal (and often, the MCU datasheet doesn't provide the necessary information to determine this anyway..!), sensitive probes to measure and verify operation, and other test equipment to evaluate EMC (e.g. RF amplifiers, signal generator, spectrum analyzer, probes and antennas, etc.). Meaningful effort is challenging, and poorly understood -- not to say there aren't those who understand how to design and test these circuits, or that it's particularly difficult when given the tools to do so, but merely that the knowledge itself is poorly distributed, few know it -- few need to know it, because crystals generally
just work -- and so, responses largely avoid discussing anything of actual consequence, any elephants in the room.
Finally, notice how 1. there is wildly insufficient information to make any kind of statement on EMC of this PCB to begin with (we see only a tiny excerpt, no idea what signals are and where they're going, how the rest of the board is laid out, what it connects to, the full schematic, and so on), and 2. no one has offered any quantitative -- not even qualitative, for that matter -- assertions about potential improvements or tradeoffs -- in terms of emissions, susceptibility, interference, jitter, you name it. Anything, that would be material to the topic, what it is you're actually accomplishing by moving things around, if at all -- and we can measure these, and while it's maybe not easy to calculate those outward effects from say board geometry alone, at least some ballpark figures can be given from experience, or by rough analysis.
So you can see how this behavior can be problematic, creating extra busy-work when none is required, and distracting from issues of importance.
---
Now, I don't write this as an admonishment to any particular posters in this thread (but, if [posters] can accept this as constructive criticism, then please, by all means-), but more as a warning to you [the OP], you will find this anti-pattern time and again in the world of human endeavors, and it will be up to you to cut through the smoke screen and realize for yourself when and where a lack of substance exists. And, I appreciate that [posters] have so accurately portrayed this behavior, and created this teachable moment.
On a similar note, I wrote this a bit ago:
https://electronics.stackexchange.com/a/705405/311631 you may find the comments on layout in general, and avoiding early optimization, and analysis paralysis, of interest.
Tim