fix: drive dbg_dk_idx_i only during full-dk readback

The core's ek_rd_addr/dkp_rd_addr muxes give dbg_dk_idx_i priority over
dbg_byte_idx_i whenever it falls in the ek (768..1567) or dk_pke (<768)
region. The wrapper drove both dbg_byte_idx_i and dbg_dk_idx_i from the
same rdb_addr, so an ek readback of the rho tail (bytes 768..799) was
hijacked to read ek[0..31] instead — corrupting the public key fed to
Encaps and causing Decaps implicit rejection (full 32-byte ss mismatch).

Present rdb_addr on dbg_dk_idx_i only when rdb_sel==2 (full-dk); hold it
at 0xFFF otherwise so ek/dk_pke readbacks aren't corrupted. Verified by
KeyGen->Encaps->Decaps round trip (MLKEM_RESULT=PASS).

rtl/ip/mlkem/top/mlkem_axi_slave.v

@@ -191,6 +191,12 @@ module mlkem_axi_slave (
     wire [7:0]  dbg_byte_o, dbg_dk_o, dbg_ct_o;
     wire [255:0] ss_o;
     wire        dbg_byte_sel = (rdb_sel == 2'd1);   // 0=ek, 1=dk_pke
+    // The core's ek/dkp read-address muxes give dbg_dk_idx_i priority over
+    // dbg_byte_idx_i whenever it lands in the ek (768..1567) or dk_pke (<768)
+    // region (mlkem_top: ek_rd_addr / dkp_rd_addr). So only drive dbg_dk_idx_i
+    // with the real index during a full-dk readback (rdb_sel==2); otherwise hold
+    // it out of range (0xFFF) so an ek/dk_pke readback isn't corrupted.
+    wire [11:0] dbg_dk_idx = (rdb_sel == 2'd2) ? rdb_addr : 12'hFFF;
 
     // one-cycle write-enable pulses into the core (combinational off the synced
     // edge; pre_sel_l/pre_addr_l/pre_byte_l are stable in sys, sampled by core).
@@ -235,7 +241,7 @@ module mlkem_axi_slave (
         .dbg_byte_idx_i (rdb_addr[10:0]),
         .dbg_byte_o     (dbg_byte_o),
         // full-dk byte readback
-        .dbg_dk_idx_i   (rdb_addr),
+        .dbg_dk_idx_i   (dbg_dk_idx),
         .dbg_dk_o       (dbg_dk_o),
         // remaining debug taps unused
         .dbg_rho_o    (/* unused */),