Emitter bugfix

Full Changelog: v0.1.1...v0.1.2

Fixed a corner-case bug in the emitter code causing an additional newline to be made.

Added a nix development environment.

new snippet syntax

* rewrite snippet parser to use different snippet open- and close
  syntax. This enables support for using snippets in languages w/o
  line comments, i.e. the comment spans multiple lines until
  explicitly ended.
  Examples are HTML ('<!-- ... -->'), CSS and classic C ('/* ... */').
  The old parser would attempt submitting the remainder of the line to
  the JSON parser for parsing, causing an error. Given explicit open-
  and end tags, the parser will disregard anything following the close
  tag.
* minor adjustments to remove warnings from mypy
* mark package as using types - this enables code consuming this library
  to benefit from the type annotations.

Rework Emitter -> Section

• Split Emitter class. Emitter now takes a Section and some buffer, writing the output of the Section into the buffer.
• Section retains the Emitter class write API, but emitln_r/emitln_d are deprecated in favor of emitln(...).indent()/emitln(...).dedent().
• All output methods in Section return the Section itself - it is a fluent API permitting something like emitln("foo").indent() instead of emitln_r("foo") which is arguably clearer to read.
• Section.add_section() now takes a section as an argument - Sections can thus be created and returned as a value to be used inside another Section unlike before.
• Section.ensure_padding_lines(n) has gotten several bugfixes.
  • If called at the start of output, it is ignored.
  • If called at the very end of output, it is ignored.
  • Padding is intended to be analogous to padding in a HTML-sense, providing spacing BETWEEN contents, not creating a margin surrounding the snippet.
  • Finally, multiple padding calls merge, meaning the highest padding requirement takes effect. Explicitly added newlines count toward the padding requirements.

v0.0.7

• emitter - ensure_padding_lines(num_lines: int) -- use to guarantee at least n empty lines of spacing between the next line and the prior line (if any). Is a no-op if called with the emitter's buffer being empty (no content written out yet). This method is useful when wanting to ensure a certain amount of spacing between paragraphs/block of code.
• gcgen.api.tree.py - NodeVisitor and NodeTransformer classes and associated on_visit and on_transform decorators. Collectively, the tree module provides utility code to traverse/visit and transform/rewrite nested object structures (~trees). This is very useful when reading in a model which needs validation (NodeVisitor's), some transformation into a more codegen-friendly format (NodeTransformer's) and finally when writing out the model to code (NodeVisitor). This approach is inspired by multi-pass compilers.

v0.0.6

• fixed bug where a snippet ending in the middle of a line would cause the end tag to be written on that same line, causing the next parse of the file to fail.
• <project_root> directory (as identified by the first directory up which contains a gcgen_project.ini file or, failing that the first directory up which has a .git folder) is now added to the front of sys.path. This means imports of code relative is done relative to the project directory.

v0.0.5

• snippets may now take a single argument which has to be a valid JSON value
• default open/close tags are <<? and ?>> instead of [[start and end]] -- many editors balance brackets, so the former default was unwieldy.

v0.0.4

missing file

0.0.3

tweak default log levels

v0.0.2

fix project url, add gh repo link to docs

first release

Initial release.

Still some rough edges, particularly as regards to pretty-printing errors on the terminal.