need to improve how we handle buffers

[kushalkolar]

I just noticed that before #143 when we initialized any GraphicFeature, the data were being copied to a new array because of astype(np.float32) which has copy=True by default: https://github.com/kushalkolar/fastplotlib/pull/143/files#diff-e9a4f655a73cc26eb39868b9b4f44940cccfa247f6c433559cec77954383a198L46

We need to think about how to properly deal with updating the data in these buffers when the arrays are read-only, such as numpy memmaps:

>>> img_graphic.data._buffer.data

memmap([[ 14.957001 ,   8.239777 ,  24.973755 , ...,   2.7485046,
           5.922577 ,   3.677765 ],
        [ 91.104004 ,  39.25052  ,   6.083435 , ...,  38.935028 ,
          57.211304 ,  23.596008 ],
        [121.86125  ,  74.614716 ,  20.688904 , ...,  25.005035 ,
          73.836426 ,  12.158752 ],
        ...,
        [ 35.85089  ,  22.899536 ,  21.668213 , ...,  93.56639  ,
          38.996155 ,  32.658997 ],
        [ 15.931152 ,  12.423279 ,  10.606384 , ...,  64.8004   ,
          21.833801 ,   8.394348 ],
        [ 54.492004 ,   6.143982 ,  18.702332 , ...,  22.917175 ,
          15.810699 ,  30.137512 ]], dtype=float32)

Trying to assign, results in the following:

>> img_graphic.data = 0
---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
Input In [44], in <cell line: 1>()
----> 1 iw_cnmf.plot.graphics[0].data[0] = 0

File ~/Insync/gdrive/repos/fastplotlib/fastplotlib/graphics/features/_data.py:114, in ImageDataFeature.__setitem__(self, key, value)
    110 def __setitem__(self, key, value):
    111     # make sure float32
    112     value = to_gpu_supported_dtype(value)
--> 114     self._buffer.data[key] = value
    115     self._update_range(key)
    117     # avoid creating dicts constantly if there are no events to handle

ValueError: assignment destination is read-only

But we are only showing a part of the entire large memmap in those cases so we could just set the texture data entirely if it's a lazy-loading type, maybe a nice way to do it would be to have something like:

import fastplotlib
from numpy import memmap
from mesmerize_core.arrays import LazyArray

fastplotlib.config.lazy_array_types.append(memmap, LazyArray)
``

And then in the various `GraphicFeatureIndexable._set()` we can set `_buffer.data` entirely if they are one of those `lazy_array_types`.

[kushalkolar]

1. Refactor indexable graphics, don't have __call__ to get the data under the buffer, make it explicit so it has to be feature.buffer.data to get the data.
2. A kwarg for isolating buffers with this type of usage:
   • common_buffer=None - all buffers are isolated from input arrays
   • common_buffer="data"
   • common_buffer=["data", "colors"] - when using non-isolated buffer for colors, an array must be provided for the colors.
3. If a common buffer is specified for any feature, then the dtype of the input array must be GPU compatible else raise

[kushalkolar]

Consider making features properties. This way we can define the set behavior better without the awkward way it's done using setattr in Graphic