feat: Gemma 4 support

[giladgd]

Description of change

• feat: Gemma 4 support
• feat: automatically enable flash attention when optimal
• feat: improve inference performance when a grammar is active
• feat: more precise resource usage estimation
• feat: resource usage capping
• feat: useMmap: "auto"
• feat: support Q1_0 quant
• feat: improve stability on unified memory systems
• feat: disabled residency sets on macOS by default for better OS responsiveness
• feat: default to progressLogs: "stderr"
• feat: optimized prebuilt binaries for arm architectures
• fix: apply llama.cpp patches if pending required PRs weren't merged yet
• fix: MXFP4_MOE quant name
• fix: Vulkan backend successful load detection even when no devices are available
• fix(CLI): avoid redownloading existing models that consist of multiple parts from a URI
• fix: optimize checkpoints management when using grammar
• fix: improve stability when loading huge models
• fix: adapt to breaking llama.cpp changes

Resolves #594
Fixes #600
Fixes #602
Resolves #596
Fixes #593

Pull-Request Checklist

• Code is up-to-date with the master branch
• npm run format to apply eslint formatting
• npm run test passes with this change
• This pull request links relevant issues as Fixes #0000
• There are new or updated unit tests validating the change
• Documentation has been updated to reflect this change
• The new commits and pull request title follow conventions explained in pull request guidelines (PRs that do not follow this convention will not be merged)

[0x7s0lt1]

🙏

[dbyale]

🙏

[giladgd]

Still working on it, hope to finish in the next few days or so.
It’s going to be a bigger change than I initially planned but will significantly improve general stability and performance.

[rafeautie]

😀

[ido-pluto]

LGTM