Use default equality for asm_syntax structs

[elazarg]

Use C++20 default equality operator for the internal eBPF syntax. This is more robust to changes, and any instruction should be encoded in a way that supports this operator.

The only struct that uses a custom operator is Call; it contains type and name information that do not really belong there, and should be extracted in a future PR (which may help generalizing callx implementation from the basic support in #584).

Some unmarshalling tests are affected:

• Unary byteswap operations have an unused is64 field; for now it is worked around in the test. (Branch cleanup-c++20 implements an alternative, encoding width for all unary operations).
• Negation was incorrectly implemented as the comment in the code suggested, and is amended accordinly.

IMO this PR is better merged before #591.

[coveralls]

coverage: 90.171% (+0.04%) from 90.13%
when pulling a695ea6 on default-equality
into 1056586 on main.

[dthaler]

• Negation was incorrectly implemented as the comment in the code suggested, and is amended accordinly.

bpf_conformance wasn't catching this. I've now filed
Alan-Jowett/bpf_conformance#204
to update bpf_conformance.