wasm2c: Fix WASM_RT_THREAD_LOCAL to check for __thread and add an error path

[shravanrn]

wasm2c's thread_local needs two fixes

1. WASM_RT_THREAD_LOCAL should use __thread compiler intrinsic where possible (gcc, clang, icc) to avoid making C11 a hard dependency to compile the wasm2c runtime.
2. The WASM_RT_THREAD_LOCAL fallback when there is no thread_local implementation either from compiler intrinsics or C11 is to #define WASM_RT_THREAD_LOCAL, i.e., leave it empty. This is pretty unsafe, as this fallback would mean wasm2c generated code would have race conditions if two instances are run in separate threads. This PR adds an extra macro (WASM_RT_ASSUME_SINGLE_THREADED_APP) for users to declare their app is single threaded, at which point this fallback is allowed.