Replace constant defines by static const typed variables

[eliemichel]

This may be a naive question, but is there any reason for these constant values to be declared as macro variables using #define rather than typed static consts?

We already use static const for enum values now so there should be no support issue.

[eliemichel]

(Will update webgpu.yml and gen script if relevant)

[kainino0x]

None of us had strong opinions on which way to go, but it's a bit more traditional. It feels too arbitrary to be worth changing, but I don't know.

• for example gl.h and vulkan_core.h do it this way
• there was some preference for macros expressed here: https://github.com/webgpu-native/webgpu-headers/pull/34
• it also turns out there are some limitations with static const that might push us toward macros: https://github.com/webgpu-native/webgpu-headers/issues/339 (though I don't think we'll use macros to define all of the flags constants - incidentally Vulkan also uses static const for 64-bit flags).

[kainino0x]

• it also turns out there are some limitations with static const that might push us toward macros: Latest header does not compile on MSVC due to const expressions #339 (though I don't think we'll use macros to define all of the flags constants - incidentally Vulkan also uses static const for 64-bit flags).

I checked, and this applies to MSVC C but not C++. The pattern this will create problems for is e.g.:

static const uint64_t kMyUsages = WGPUBufferUsage_CopyDst | WGPUBufferUsage_Uniform;

but users can deal with this, by themselves using a macro:

#define MY_USAGES (WGPUBufferUsage_CopyDst | WGPUBufferUsage_Uniform)

[kainino0x]

I found a useful series of answers about C here: https://stackoverflow.com/questions/1674032/static-const-vs-define-vs-enum (which was harder to find than the equivalent C++ question here: https://stackoverflow.com/questions/1637332/static-const-vs-define)

Given we probably don't get any significant type safety out of these (they're mostly uint32_t and they're no worse than UINT32_MAX etc.) I would lean weakly toward just keeping the #defines.