Exactly this. There are no literals in C that create composite types. There are no composite types inherent to the language. All these types are defined in (system) includes.
And zero-terminated strings are not strictly worse than other length-prefixed string forms. They save some space -- sure, less relevant today -- as well as provide an in-band termination signal -- which is hacky, again sure, but it is convenient when looking at a hex dump for example.
Because a float contains bit fields including sign of the exponent, sign, mantissa, sign of the mantissa. It's a bit of a pedantic argument but technically it makes sense.
You could call them composite in that sense, but in C, composite types are types that are composed of other C types (structs, unions, arrays, ... functions? Not 100% sure of the specifics).
Also, the representations of floats and doubles isn't precisely specified, at least IEEE754 is not a strict requirement (not sure about the technical implications from what's actually specified).
I'm not really sure I get why the distinction between composite and non-composite types is important if the only difference is that you can't easily access sub-parts of the non-composite types.
I guess we can cook up some arguments about language purity and orthogonality. Introducing literals that create composite types might indeed create some difficulties for existing compilers that want to follow the latest standard.
It requires the compiler to be able to create types even before parsing, to have composite types without any source code location... you'd have to check with a specific compiler but it's not hard to imagine some potential issues.
There's also outward facing changes that introduce complexity... new literal types (+ new syntax) are required because otherwise old could would break (sizeof...) etc. etc. Existing parsers and tooling will have to be adapted, too...
Generally it's nothing that should be insurmountable, but why accept any of that if there's nothing to gain except a feature that is requested mainly by detractors of the ecosystem, a feature that goes squarely against the grain of the language...?
Any language is proud to be able to implement its features as library components around its core capabilities. Why introduce a new C type as a structure that can easily be defined as an aggregate type in source code?
And zero-terminated strings are not strictly worse than other length-prefixed string forms. They save some space -- sure, less relevant today -- as well as provide an in-band termination signal -- which is hacky, again sure, but it is convenient when looking at a hex dump for example.