wt_icache: don't fallback to the TLB_MISS state when stalled in READ

Currently, when there is an I$ miss but the icache is locked in the READ
state, it would go in the TLB_MISS in the next cycle, which is useless.

This fixes this by setting the next state to READ when `stall_req_i' is
set in the case of a cache miss.

Signed-off-by: Alban Gruin <alban.gruin@irit.fr>

src/cache_subsystem/wt_icache.sv

@@ -260,6 +260,8 @@ end else begin : gen_piton_offset
                 miss_o         = ~paddr_is_nc;
                 state_d        = MISS;
               end
+            end else if (stall_req_i) begin
+              state_d = READ;
             end
           // bail out if this request is being killed (and we missed on the TLB)
           end else if (dreq_i.kill_s2 || flush_d) begin