fix: refactor image load and add i/dcache flush ops

[liulog]

For situations requiring a virtual machine restart: Due to residual guest data in memory/cache, guest instances that are not starting for the first time may be affected by this residual data.

When the system is running, it is necessary to clear the memory allocated to the virtual machine. At the same time, before startup, the kernel image and device tree need to be flushed to memory to avoid errors caused by fetching instructions directly from memory without caching in the early stages of virtual machine startup.

In addition, the cpu cores allocated to virtual machines should ideally perform TLB and cache cleanup operations locally to prevent the virtual machine from being affected by dirty data after the MMU is enabled or the cache is enabled. This should be done by the Hypervisor(hvisor), so changes to this part will be described in a subsequent hvisor PR.

[liulog]

Add one json filed, setting need_clear to true will execute clearing operations.

For example:

[liulog]

Testing revealed that performing a flush operation after memset 0 or binding hvisor-tool to a specific CPU core both achieve the desired result.

This means either binding to a CPU core or clearing followed by a flush. The ultimate goal is to ensure that the data in memory is clean and undisturbed.

The flush operation is used here to ensure that the clear 0 operation is also written in memory, so that any CPU executing the load image will see a clean memory space.