Incorrect IRQ priority with LCOF

[JJ-Gaisler]

I have a test case which sets all software triggerable bits in MIP.
For Spike with H and Sscofpmf this should mean the MIP value 0x20aa.
It does however look like Spike doesn't prioritize the LCOF interrupt correctly since it is considered a custom interrupt.

Prior to LCOF and SGEI, MEI was the highest priority standard interrupt, but it looks like the following code hasn't been updated to deal with that.

// nonstandard interrupts have highest priority
if (enabled_interrupts >> (IRQ_M_EXT + 1))
  enabled_interrupts = enabled_interrupts >> (IRQ_M_EXT + 1) << (IRQ_M_EXT + 1);

Tested the following and that seems to have been enough.

// nonstandard interrupts have highest priority
if (enabled_interrupts >> (IRQ_LCOF + 1))
  enabled_interrupts = enabled_interrupts >> (IRQ_LCOF + 1) << (IRQ_LCOF + 1);
// standard interrupt priority is MEI, MSI, MTI, SEI, SSI, STI, LCOF

https://github.com/riscv-software-src/riscv-isa-sim/blob/f7d0dba6012d49cb75b2f175a31a85cf4b2db3b3/riscv/processor.cc#L697C32-L697C41

Reproducible:
hypervisor_tests.txt
spike -m0x80000000:0x10000000,0x20000000:0x1000,0x4000:0x1000,0xB000:0x1000 --isa rv64gvh_svinval_zba_zbb_zbc_zbs_zfh_zbkb_zicsr_zifencei_sscofpmf_smepmp_zicbom_zicntr_zihpm_svadu_sstc_zicond -p1 hypervisor_tests.txt

[demin-han]

PR #1830 would solve this.