Lna investigations and fix (#3240)

* The LNA gain field 'L' lives in LPF_VGA_2 (reg 6), not RXRF_2 (reg 2). RXRF_2 only holds LNAgain_SPI_EN. The configure_rx_gain() function wrote 'L' into lpf_vga_2 but marked RXRF_2 dirty => the LNA value only got flushed incidentally because the following VGA write marks LPF_VGA_2 dirty. Drop the stray RXRF_2 dirty flag and rely on the LPF_VGA_2 flag, which carries both L and VGA.

* The baseband build defined IS_H1_R9 as the compile-time constant 0, on the (now incorrect) assumption that detected_platform() isn't linked. The sd_over_usb image is the only baseband target that compiles the IS_H1_R9-gated upstream files (rf_path.c, si5351c.c, sgpio.c, platform_gpio.c, platform_scu.c), and it *does* link platform_detect.c and call detect_hardware_platform() at startup. Forcing IS_H1_R9=0 dead-stripped every 'if (IS_H1_R9)' branch, so a real HackRF One r9 was set up with the OG clock/GPIO/RF-path config in that image.
Define IS_H1_R9 as the runtime '(detected_platform()==BOARD_ID_HACKRF1_R9)', mirroring upstream platform-detect.cmake for HACKRF_ONE. All usages are '#ifdef IS_H1_R9' + 'if (IS_H1_R9)' (no arithmetic '#if'), so a function-call expression is safe, single-quoted -D with no internal spaces keeps it intact through the shell-split DDEFS blob.

---------

Co-authored-by: gullradriel <gullradriel@users.noreply.github.com>

firmware/application/hw/max2839.cpp

@@ -307,9 +307,8 @@ void MAX2839::configure_rx_gain() {
     }
 
     _map.r.lpf_vga_2.L = lna::gain_ordinal(lna_gain);
-    _dirty[Register::RXRF_2] = 1;
     _map.r.lpf_vga_2.VGA = vga::gain_ordinal(vga_gain);
-    _dirty[Register::LPF_VGA_2] = 1;
+    _dirty[Register::LPF_VGA_2] = 1; /* both L (LNA) and VGA live in LPF_VGA_2 */
     flush();
 }