ZynqMP ZCU102 SD-card Linux boot fixes

Fixes for booting PetaLinux 2025.2 from SD card on ZCU102:
- hal/zynq: implement hal_dts_fixup() for DTB bootargs override
- hal/zynq: add hal_get_timer_us() using ARMv8 generic timer
- src/sdhci: add udelay settle times and CMD0 retry loop for cold boot
- src/boot_aarch64: clean D-cache and disable MMU/caches at EL2 before
  jumping to Linux (ARM64 boot protocol requirement)
- config/examples/zynqmp_sdcard: reduce DISK_BLOCK_SIZE from 512KB to
  128KB to avoid SDMA boundary-crossing corruption

Author: dgarske

config/examples/zynqmp_sdcard.config

@@ -17,7 +17,7 @@ SIGN?=RSA4096
 HASH?=SHA3
 IMAGE_HEADER_SIZE?=1024
 
-# Debug options
+# Debug options - for production, set DEBUG=0 and drop DEBUG_ZYNQ
 DEBUG?=0
 DEBUG_SYMBOLS=1
 DEBUG_UART=1
@@ -36,9 +36,6 @@ CFLAGS_EXTRA+=-DSDHCI_FORCE_CARD_DETECT
 EXT_FLASH?=0
 NO_XIP=1
 
-# ELF loading support
-ELF?=1
-
 # Boot Exception Level: leave wolfBoot at EL2 for handoff to Linux (matches
 # the standard PetaLinux U-Boot flow and preserves KVM/hypervisor use of
 # EL2). The EL2 Linux-cleanup path in do_boot() will clean dcache/disable
@@ -58,6 +55,8 @@ RAM_CODE?=0
 DUALBANK_SWAP?=0
 PKA?=0
 WOLFTPM?=0
+ELF?=1
+#DEBUG_ELF?=1
 
 # Toolchain
 USE_GCC=1
@@ -79,8 +78,10 @@ WOLFBOOT_NO_PARTITIONS=1
 CFLAGS_EXTRA+=-DBOOT_PART_A=1
 CFLAGS_EXTRA+=-DBOOT_PART_B=2
 
-# Disk read chunk size (512KB)
-CFLAGS_EXTRA+=-DDISK_BLOCK_SIZE=0x80000
+# Disk read chunk size. Must be less than the SDHCI SDMA buffer boundary
+# (512KB default, SDHCI_DMA_BUFF_BOUNDARY in sdhci.h). Using 512KB causes
+# a boundary-crossing edge case that corrupts reads under optimization.
+CFLAGS_EXTRA+=-DDISK_BLOCK_SIZE=0x20000
 
 # Linux rootfs is on partition 4. Device naming depends on whether both
 # ZynqMP SDHCI controllers are enabled in the XSA / device tree: