This PR contains replacement for the viminfo file discussed in #999.

• ShaDa file dumping.
• Merging.
• Writing.
  • Registers.
  • History item.
  • Last search/substitute pattern.
  • Last substitute replacement string.
  • Variable values.
  • Global marks.
  • Local marks.
  • Jump list.
  • Buffer list.
  • Buffer change list.
• Reading.
  • Registers.
  • History item.
  • Last search/substitute pattern.
  • Last substitute replacement string.
  • Variable values.
  • Global marks.
  • Local marks.
  • Jump list.
  • Buffer list.
  • Buffer change list.
  • v:old_files list (is determined based on marks in shada, so has no equivalent in “writing section”).
• Functional tests.
  • Registers.
  • History item.
  • Last search/substitute pattern.
  • Last substitute replacement string.
  • Global marks.
  • Local marks.
  • Jump list.
  • Buffer list.
  • Buffer change list.
  • Compatibility support:
    • Ignoring global marks with unknown names when reading.
    • Ignoring local marks with unknown names when reading.
    • Ignoring registers with unknown names when reading.
    • Ignoring registers with unknown types when reading.
    • Ignoring history with unknown type when reading.
    • Preserving global marks with unknown names when merging.
    • Preserving local marks with unknown names when merging.
    • Preserving registers with unknown names when merging.
    • Preserving registers with unknown types when merging.
    • Preserving history with unknown type when merging.
    • Preserving items with unknown type when merging.
    • Preserving additional_data/additional_elements in
      • Registers.
      • History items.
      • Last search/substitute patterns.
      • Last substitute replacement string.
      • Global marks.
      • Local marks.
      • Jump list.
      • Buffer list.
      • Buffer change list.
  • Variable values.
  • v:old_files list.
  • Merging:
    • History merging:
      • Ignoring read history item if NeoVim instance already contains its own with ge timestamp when reading.
      • Using read history item if NeoVim instance already contains its own with lt timestamp when reading.
      • Ignoring read history item if NeoVim instance already contains its own with ge timestamp when writing.
      • Using read history item if NeoVim instance already contains its own with lt timestamp when writing.
      • Reading history correctly with messed up timestamp order.
      • Writing history correctly with messed up timestamp order.
    • Replacing last search/substitute pattern with an older one when reading.
    • Replacing last search/substitute pattern with read one when reading with bang.
    • Replacing last search/substitute pattern with an older one when writing.
    • Replacing replacement string with an older one when reading.
    • Replacing replacement string with read one when reading with bang.
    • Replacing replacement string with an older one when writing.
    • Merging jump list when reading.
    • Merging jump list when writing.
    • Merging change list when reading.
    • Merging change list when writing.
    • Replacing global mark with an older one when reading.
    • Replacing global mark with read one when reading with bang.
    • Replacing global mark with an older one when writing.
    • Replacing local mark with an older one when reading.
    • Replacing local mark with read one when reading with bang.
    • Replacing local mark with an older one when writing.
    • Replacing register with an older one when reading.
    • Replacing register with read one when reading with bang.
    • Replacing register with an older one when writing.
  • Ignoring existing file when writing with bang.
  • Converting some data from &encoding when writing:
    • Converting register values when writing.
    • Converting variable values when writing.
    • Converting history strings when writing.
    • Converting last used search/substitute patterns when writing.
    • Converting last used replacement strings when writing.
  • Converting some data to &encoding when reading:
    • Converting register values when writing.
    • Converting variable values when writing.
    • Converting history strings when writing.
    • Converting last used search/substitute patterns when writing.
    • Converting last used replacement strings when writing.
  • Reaction on various errors in ShaDa file:
    • Items with invalid header.
    • Registers.
    • History items.
    • Last search/substitute patterns.
    • Last substitute replacement string.
    • Global marks.
    • Local marks.
    • Jump list.
    • Buffer list.
    • Buffer change list.
    • Variables.
  • s item size limit.
  • s item size limit being used when merging for “foreign” items.
  • Using .tmp.b when .tmp.a already exists.
  • Using .tmp.z when .tmp.a….tmp.x already exist.
  • Leaving temporary file when there was error in target file.
  • Failing when .tmp.a….tmp.z files exist.
  • Positive fixnum timestamp being read correctly.
  • Uint8 timestamp being read correctly.
  • Uint16 timestamp being read correctly.
  • Uint32 timestamp being read correctly.
  • Uint64 timestamp being read correctly.
• Documentation update.
• Converting some data from &encoding when writing.
• Converting some data to &encoding when reading.
• Safer variant of dumping viminfo (dump to temporary file then rename).
• Removing old viminfo code (currently masked with #if 0 in a ShaDa file for
  reference).
• Fixing linter errors.
• Fixing existing failing tests.