Consider eliding reader produced metadata

[lread]

Adapated from lread/rewrite-cljc-playground#63, fix to be reapplied to rewrite-clj

Problem/Opportunity
Rewrite-clj and therefore rewrite-cljc automatically coerces metadata to nodes.

When the metadata comes from the reader rather than the actual source code, this might cause confusion (see #37).

For example in Clojure, :line and :column metadata are added to quoted lists:

> clj
Clojure 1.10.1
user=> (meta '(1 2 3))
{:line 1, :column 8}

This metadata is not in the original source code. It is added by the reader.

A more obscure use case is under sci (and therefore also babashka).
Sci adds positional metadata :line, :column, :end-line and :end-column to all elements that accept metadata.

> bb --repl
Babashka v0.2.3 REPL.
Use :repl/quit or :repl/exit to quit the REPL.
Clojure rocks, Bash reaches.

user=> (meta '(1 2 3))
{:line 1, :column 7, :end-line 1, :end-column 15}
user=> (meta [4 5 6])
{:line 3, :column 7, :end-line 3, :end-column 14}
user=> (meta #{1 2 3})
{:line 5, :column 7, :end-line 5, :end-column 15}
user=> 

For rewrite-cljc, this metadata manifests itself problematically during coercion from Clojure form to rewrite-cljc node.

Proposed Solution
I think that suppressing well known reader metadata by default would be less surprising to users of rewrite-cljc.

This is technically a breaking change in default behavior. I can only imagine existing rewrite-clj users who have stumbled against this behavior would have special code to work around it.

If a user of rewrite-cljc, for some reason, felt a need to preserve the existing rewrite-clj behavior, we could support that with a dynamic variable. Perhaps rewrite-cljc/*elide-metadata* would default to [:line :column :end-line :end-column] but could be overridden to nil to offer a route to backwards compatibility. But on second thought, I think we'll just hardcode the eliding; the behaviour of including it, can be rationalize as a bug.

This change would:

1. help mainstream users of rewrite-cljc under Clojure avoid an oddity in handling of quoted lists.
2. simplify my sci tests; it would allow me to abandon my sci fork (which compensates from some metadata handling) and use sci master in a more straightforward manner.

Alternative Solutions
None, at present.

Action
I will experiment with above idea, and if fruitful, will effect the change.