New argument for implicit_assignment_linter allowing lazy assignment

[MichaelChirico]

Related but I'm not sure it's the same as #2015.

Another type of violation to implicit_assignment_linter() that I find a bit awkward to refactor looks like:

if (rare_condition1 && rare_condition2 && foo(key <- val)) {
  bar(key)
}

Lazy assignment means it's rare that key is actually assigned, so this refactor is wasteful:

key <- val
if (rare_condition1 && rare_condition2 && foo(key)) {
  bar(key)
}

The version that seems preferred gets close to an unnecessary_nested_if() violation:

if (rare_condition1 && rare_condition2) {
  key <- val
  if (foo(key)) {
    bar(key)
  }
}

This increases the indentation vs. the original version; this problem gets worse (possibly combinatorially) if there's more than one such assignment in the same if() condition.

IINM this is a subset of #2015, but still could be used in different situations. Again there's no such carveout in the style guide, so this usage should be disallowed by default.

[AshesITR]

Wasteful in the sense that key can't be gc() ed immediately after the if branch?

[MichaelChirico]

more importantly we could have key <- costly_call() that can be avoided