`range` is not handling `InexactError`

[sadish-d]

range calls Base._linspace, which calls round which is throwing an InexactError for the following case:

start = 0.7071067811865475
len = 1449
range(start, nextfloat(start), len) # works
Base._linspace(start, nextfloat(start), len) # works
range(start, nextfloat(start), len + 1) # fails
Base._linspace(start, nextfloat(start), len + 1) # fails

The variable len is shadowed inside _linspace by one that is of type Int or higher.

And the type of this len is used by the round function to determine what format to round to.

In my machine, Int is Int64. So round tries to convert some large float to Int64 and fails.

This works:

range(start, nextfloat(start), Int128(len + 1))
Base._linspace(start, nextfloat(start), Int128(len + 1))

My hack right now is to define my own range function:

my_own_range_function(start, stop, len::Integer) = Base.range(start, stop, Int128(len))

I'm not sure hard-coding Int128 will handle all cases where round throws an error.

related: #57113