benchmark: remove duplicate ffi add-64 benchmark

[soulee-dev]

benchmark/ffi/add-64.js and benchmark/ffi/add-f64.js measured the same
thing: both dlopen the fixture library, bind add_f64 with the signature
{ return: 'f64', arguments: ['f64', 'f64'] }, and call it n = 1e7 times.
The only difference was the argument constants (add(20.5, 21.5) vs
add(1.5, 2.5)), which does not change what is being measured.

add-f64.js came first (#62918). add-64.js was added later in #63068 as
part of a batch of add-* benchmarks; git records it as a copy of
add-f64.js with only the constants changed to that batch's convention, and
nothing in the review of either PR treats it as a distinct case. Nothing in
benchmark/, doc/, test/ or tools/ references the add-64 name.

This removes add-64.js and keeps add-f64.js, whose name matches the
add_f64 symbol it measures and the naming of the other add-<type>.js
files.

Verified locally:

$ NODE=$PWD/out/Release/node make lint-js
$ out/Release/node benchmark/ffi/add-f64.js n=1000
ffi/add-f64.js n=1000: 6,655,574.04
$ python3 tools/test.py -j 4 ffi
[00:02|% 100|+  13|-   0]: Done

Refs: #63068
Refs: #62918

[nodejs-github-bot]

Review requested:

• @nodejs/performance