Consider using a single value to represent an invalid ID

[lmb]

The runtime currently reserves two values to mean "invalid ID": 0 and EBPF_ID_NONE (which is the same as MAX_UINT32). This leads to duplicate checks / inconsistencies in code checking IDs for validity.

For reference, Linux uses 0 as the invalid ID. MAX_UINT32 is a valid ID, afaik.

@dthaler suggested that it might make sense to define EBPF_ID_NONE to be 0.