add Loongarch init support

[yzewei]

What was done

Completed a single-core (1 vCPU) guest VM for LoongArch64:

1. Completed the basic integration path of LoongArch64 in arch/devices/vmm:

• LoongArch startup/register initialization, FDT generation, MMIO/interrupt chip integration, memory layout and loading process adaptation.

• Completed LoongArch branches for critical paths such as builder, vstate, and mmio, ensuring the VM can normally enter the guest and run test cases.

2. Completed test and runtime chain adaptation:

• Supported the LoongArch guest-agent running mode on the test side (including runtime dependency handling), and fixed console startup timing issues.

• Improved relevant test cases under LoongArch.

Tests Passed

Integration Tests (LoongArch, Single Core)

• configure-vm-1cpu-256MiB

• configure-vm-1cpu-1GiB

• vsock-guest-connect

• tsi-tcp-guest-connect

• tsi-tcp-guest-listen

• multiport-console

Unit Tests

• All unit tests in vmm passed.

Future Improvements

• SMP (2+ vCPUs): Currently, due to the lack of an upstream version of the interrupt mechanism (IPI) on both the host KVM and guest kernels, startup with 2 vCPUs fails under SMP.

About LoongArch

The LoongArch architecture (LoongArch) is an Instruction Set Architecture (ISA) that has a RISC style.

Documentations:
ISA:
https://loongson.github.io/LoongArch-Documentation/LoongArch-Vol1-EN.html
ABI:
https://loongson.github.io/LoongArch-Documentation/LoongArch-ELF-ABI-EN.html
More docs can be found at:
https://loongson.github.io/LoongArch-Documentation/README-EN.html

[jakecorrenti]

/gemini review

[gemini-code-assist]

Code Review

This pull request introduces support for the loongarch64 architecture, encompassing core architecture definitions, EFI system table setup, IOCSR emulation, and a specialized KVM interrupt chip. The implementation integrates LoongArch into the microVM builder and updates libkrun to manage architecture-specific kernel loading and vCPU constraints. Review feedback identifies a critical register mapping error regarding the Linux boot protocol, logic errors in IOCSR mailbox indexing, and the need to manually assert interrupts since irqfd is disabled for this architecture. Additionally, error handling improvements are suggested to replace unwrap() calls and provide more accurate error variants.