reverse backtrace-reversion

[pfitzseb]

also better rendering of constructors in backtraces.

Before:

After:

Fixes JunoLab/atom-julia-client#374.

[MikeInnes]

The current is a better default. Most users are not going to be working on base code, so the user code should come up first.

[pfitzseb]

Hm, fair enough. But there's merit to following Base's/the REPL's way of doing things as well:

julia> f(0,0)
ERROR: ArgumentError: invalid rational: zero(Int64)//zero(Int64)
Stacktrace:
 [1] Rational{Int64}(::Int64, ::Int64) at ./rational.jl:13
 [2] f(::Int64, ::Int64) at ./REPL[3]:1

The reasons for that (you only see the bottom part of long backtraces) are obviously not applicable to Juno, so yeah... Not sure if I really care about this enough to find reasons for one or the other :D