fix(ICache,ITLB): also flush itlb pipe when prefetchPipe s1_flush

[ngc7331]

ITLB does not store gpaddr to save resources, instead it takes gpaddr from L2TLB when gpf occurs, which poses a two-option requirement for the requestor (i.e. IPrefetchPipe):

1. resend the same itlb.io.req.vaddr until itlb.io.resp.miss is pulled down
2. flush gpf entry in ITLB by pulling up itlb.io.flushPipe

Otherwise, ITLB is unable to handle the next gpf and the core hangs.

However, the first point cannot be guaranteed during the speculative execution, as IPrefetchPipe sends request to ITLB at s0 stage and may receive a flush request from BPU s3 stage, IFU or Backend at s1 stage, then the same vaddr is never resend to ITLB.

Therefore, we must ensure that ITLB is flushed synchronously when IPrefetchPipe s1 stage is flushed, thus satisfying the second point. This PR implements this.

[XiangShanRobot]

[Generated by IPC robot]
commit: 60dd0f6

commit	astar	copy_and_run	coremark	gcc	gromacs	lbm	linux	mcf	microbench	milc	namd	povray	wrf	xalancbmk
60dd0f6	1.959	0.450	2.686	1.229	2.832	2.461	2.393	0.919	1.407	1.992	3.435	2.704	2.383	3.261

master branch:

commit	astar	copy_and_run	coremark	gcc	gromacs	lbm	linux	mcf	microbench	milc	namd	povray	wrf	xalancbmk
7d20eb3	1.959	0.450	2.686	1.226	2.832	2.461	2.393	0.919	1.407	1.992	3.435	2.709	2.383	3.261
35477a7
493f6e1	1.959	0.450	2.686	1.224	2.832	2.461	2.393	0.919	1.407	1.992	3.435	2.709	2.383	3.261
a60d667
7876e67

[good-circle]

Already passed tests in simulation accelerator

[eastonman]

LGTM