$(a: float) now works consistently in nim js, avoiding printing floats as ints

[timotheecour]

/cc @Araq

• fix Inconsistent float formatting in javascript. #7605
• fix var a = 2.0; echo a prints 2 in nim js timotheecour/Nim#133 (has more examples; didn't know about Inconsistent float formatting in javascript. #7605 when i wrote it)
• avoids changing the test for tpegs.nim in StringStream & more stdlib modules support for JS/NimScript #14095 which introduced js support for pegs
• after PR: $(a: float) now works more consistently and matches behavior of c, vm targets

echo 2.0 # 2.0
var b = 2.0
echo b # before PR: 2, after PR: 2.0

• it would be nice to merge $ now works for unsigned intergers with nim js #14122 before this PR so I can avoid creating a new tests/js/tdollar_float.nim

• once this is green-lighted but before merge, I'll use the single argument applyFormat in jsgen (as I did for mFloatToStr) everywhere appropriate to make code DRY

after PR

I'm thinking to introduce a compiler flag (eg --ecmascript:es5) to let users choose which js version they want to target. The Number.isInteger I'm using in this PR is available on all browsers but not on IE, but there is a polyfill we can easily introduce for that, say in std/jspolyfills. The workaround implementation should be used or not depending on the desired js version in --ecmascript; BigInt could for example be introduced for int for --ecmascript:es2020

ci failures unrelated

[metagn]

#7605 is the same as the issue in your repo

[timotheecour]

thanks! updated the PR msg; this PR will allow you to not change the test (or, better, to increase test coverage by using both float+int) in tpegs.nim in your PR #4247