Fix: hardcode GD32VW553 memory map

[ALTracer]

Detailed description

• No new features.
• The existing problem is incorrect memory map introduced by me in Feature: detect GD32VW553 #2186.
• This PR solves it by hardcoding fixed sizes of SRAM and Flash for GD32VW553. No flash driver yet.

Tested on "LCSC-boards LCKFB-GD32VW553-HMQ6", now that it arrived, via blackpill-f411ce.

Apparently 0x1ffff7e0 reads as 0xFF, so normal autodetection code gets 65536 KiB for each, this is wrong.
User Manual and Datasheet state that there are either 4 MiB or 2 MiB of SiP flash in these parts, and 320 KiB of SRAM (last 32 KiB are "shared" mysteriously, with Wi-Fi or BLE maybe). Let's just go by that. Flash is readable.

Before:

ID code 0x10307a6d: Nuclei Systems RISC-V DTM.
ID code 0x790007a3: Gigadevice BSD.
RISC-V debug v0.13/v1.0 DMI
RISC-V debug v0.13 DM
Hart has 4 data registers and 2 progbuf registers
...
Attempting 32-bit read on misa
Hart 0: 32-bit RISC-V (arch = 00000307), rv32igcbp ISA (exts = 0010912f), vendor = a36, impl = 30700
Hart has 8 trigger slots available
-> riscv32_probe
Stub for detection of GD32VW553. DBG_ID=0x418, RAM=65535, flash=65535
***  1   GD32VW5 rv32igcbp
RAM   Start: 0x20000000 length = 0x3fffc00
RAM   Start: 0x08000000 length = 0x3fffc00

After:

ID code 0x10307a6d: Nuclei Systems RISC-V DTM.
ID code 0x790007a3: Gigadevice BSD.
RISC-V debug v0.13/v1.0 DMI
RISC-V debug v0.13 DM
Hart has 4 data registers and 2 progbuf registers
...
Attempting 32-bit read on misa
Hart 0: 32-bit RISC-V (arch = 00000307), rv32igcbp ISA (exts = 0010912f), vendor = a36, impl = 30700
Hart has 8 trigger slots available
-> riscv32_probe
Stub for detection of GD32VW553. DBG_ID=0x418, RAM=320, flash=4096
***  1   GD32VW5 rv32igcbp
RAM   Start: 0x20000000 length = 0x50000
RAM   Start: 0x08000000 length = 0x400000

Your checklist for this pull request

• I've read the Code of Conduct
• I've read the guidelines for contributing to this repository
• It builds for hardware native (see Building the firmware)
• It builds as BMDA (see Building the BMDA)
• I've tested it to the best of my ability
• My commit messages provide a useful short description of what the commits do

Closing issues

[dragonmux]

Paired with what you've shared on Discord this seems reasonable - just needs a small clean-up and then we'll be happy to merge it. Shame that using the FLASHSIZE register won't work! But such are the realities of parts not setting things up right and such..

With the small tweak made, we'll be happy to merge this.