`Any.dynamic_apply` cannot be called with a lambda

[fridis]

Take this exmaple that is a based on the code given in #5896:

sum(values Sequence Any)
=>
  values.map a->
          a.dynamic_apply T,v->
            if      T : i32 then v.as_f64
            else if T : f64 then v.val
                            else compile_time_panic
        .foldf 0.0 (+)

say <| sum [3.14, 32, 16.0, 8]

It creates two errors:

 > ../build/bin/fz ../dyn_apply.fz

/home/fridi/fuzion/work/webserver/../dyn_apply.fz:4:27: error 1: Failed to infer actual type parameters
          a.dynamic_apply T,v->
            if      T : i32 then v.as_f64
            else if T : f64 then v.val
                            else compile_time_panic

In call to 'Typed_Function', no actual type parameters are given and inference of the type parameters failed.
Expected type parameters: 'R, A...'
Type inference failed for one type parameter 'R'


/home/fridi/fuzion/work/webserver/../dyn_apply.fz:4:27: error 2: Incompatible types when passing argument in a call
          a.dynamic_apply T,v->
            if      T : i32 then v.as_f64
            else if T : f64 then v.val
                            else compile_time_panic

Actual type for argument #1 'f' does not match expected type.
In call to          : 'Any.dynamic_apply'
expected formal type: 'Typed_Function R'
actual type found   : 'sum.λ.call.this.λ'
assignable to       : '**error**',
                      'sum.λ.call.this.λ'
for value assigned  : 
T,v->
            if      T : i32 then v.as_f64
            else if T : f64 then v.val
                            else compile_time_panic

To solve this, you could change the type of the target 'f' to 'sum.λ.call.this.λ' or convert the type of the assigned value to 'Typed_Function R'.

2 errors.

The second error should be suppressed since it is a result only of the previous error. But the first error should be fixed by supporting type inference for R.